id
large_string | submitter
large_string | authors
large_string | title
large_string | comments
large_string | journal-ref
large_string | doi
large_string | report-no
large_string | categories
large_string | license
large_string | abstract
large_string | versions
large list | update_date
timestamp[ms] | authors_parsed
large list | prompt
large_string | label
string | prob
float64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2303.14979 | Ning Wu | Houxing Ren, Linjun Shou, Jian Pei, Ning Wu, Ming Gong, Daxin Jiang | Lexicon-Enhanced Self-Supervised Training for Multilingual Dense
Retrieval | EMNLP 2022 Findings | null | null | null | cs.IR | http://creativecommons.org/licenses/by/4.0/ | Recent multilingual pre-trained models have shown better performance in
various multilingual tasks. However, these models perform poorly on
multilingual retrieval tasks due to lacking multilingual training data. In this
paper, we propose to mine and generate self-supervised training data based on a
large-scale unlabeled corpus. We carefully design a mining method which
combines the sparse and dense models to mine the relevance of unlabeled queries
and passages. And we introduce a query generator to generate more queries in
target languages for unlabeled passages. Through extensive experiments on Mr.
TYDI dataset and an industrial dataset from a commercial search engine, we
demonstrate that our method performs better than baselines based on various
pre-trained multilingual models. Our method even achieves on-par performance
with the supervised method on the latter dataset.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:22:21 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Ren",
"Houxing",
""
],
[
"Shou",
"Linjun",
""
],
[
"Pei",
"Jian",
""
],
[
"Wu",
"Ning",
""
],
[
"Gong",
"Ming",
""
],
[
"Jiang",
"Daxin",
""
]
] | TITLE: Lexicon-Enhanced Self-Supervised Training for Multilingual Dense
Retrieval
ABSTRACT: Recent multilingual pre-trained models have shown better performance in
various multilingual tasks. However, these models perform poorly on
multilingual retrieval tasks due to lacking multilingual training data. In this
paper, we propose to mine and generate self-supervised training data based on a
large-scale unlabeled corpus. We carefully design a mining method which
combines the sparse and dense models to mine the relevance of unlabeled queries
and passages. And we introduce a query generator to generate more queries in
target languages for unlabeled passages. Through extensive experiments on Mr.
TYDI dataset and an industrial dataset from a commercial search engine, we
demonstrate that our method performs better than baselines based on various
pre-trained multilingual models. Our method even achieves on-par performance
with the supervised method on the latter dataset.
| no_new_dataset | 0.909063 |
2303.14991 | Ning Wu | Houxing Ren, Linjun Shou, Ning Wu, Ming Gong, Daxin Jiang | Empowering Dual-Encoder with Query Generator for Cross-Lingual Dense
Retrieval | EMNLP 2022 main conference | null | null | null | cs.IR | http://creativecommons.org/licenses/by/4.0/ | In monolingual dense retrieval, lots of works focus on how to distill
knowledge from cross-encoder re-ranker to dual-encoder retriever and these
methods achieve better performance due to the effectiveness of cross-encoder
re-ranker. However, we find that the performance of the cross-encoder re-ranker
is heavily influenced by the number of training samples and the quality of
negative samples, which is hard to obtain in the cross-lingual setting. In this
paper, we propose to use a query generator as the teacher in the cross-lingual
setting, which is less dependent on enough training samples and high-quality
negative samples. In addition to traditional knowledge distillation, we further
propose a novel enhancement method, which uses the query generator to help the
dual-encoder align queries from different languages, but does not need any
additional parallel sentences. The experimental results show that our method
outperforms the state-of-the-art methods on two benchmark datasets.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:34:42 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Ren",
"Houxing",
""
],
[
"Shou",
"Linjun",
""
],
[
"Wu",
"Ning",
""
],
[
"Gong",
"Ming",
""
],
[
"Jiang",
"Daxin",
""
]
] | TITLE: Empowering Dual-Encoder with Query Generator for Cross-Lingual Dense
Retrieval
ABSTRACT: In monolingual dense retrieval, lots of works focus on how to distill
knowledge from cross-encoder re-ranker to dual-encoder retriever and these
methods achieve better performance due to the effectiveness of cross-encoder
re-ranker. However, we find that the performance of the cross-encoder re-ranker
is heavily influenced by the number of training samples and the quality of
negative samples, which is hard to obtain in the cross-lingual setting. In this
paper, we propose to use a query generator as the teacher in the cross-lingual
setting, which is less dependent on enough training samples and high-quality
negative samples. In addition to traditional knowledge distillation, we further
propose a novel enhancement method, which uses the query generator to help the
dual-encoder align queries from different languages, but does not need any
additional parallel sentences. The experimental results show that our method
outperforms the state-of-the-art methods on two benchmark datasets.
| no_new_dataset | 0.945851 |
2303.14996 | Xin-Jian Xu | Xin-Jian Xu, Chong Deng, Li-Jie Zhang | Hyperlink prediction via local random walks and Jensen-Shannon
divergence | IoP Latex, 15 pages, 1 figures | J. Stat. Mech. (2023) 033402 | 10.1088/1742-5468/acc31e | null | cs.SI cond-mat.dis-nn | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Many real-world systems involving higher-order interactions can be modeled by
hypergraphs, where vertices represent the systemic units and hyperedges
describe the interactions among them. In this paper, we focus on the problem of
hyperlink prediction which aims at inferring missing hyperlinks based on
observed hyperlinks. We propose three similarity indices for hyperlink
prediction based on local random walks and Jensen-Shannon divergence. Numerical
experiments show that the proposed indices outperform the state-of-the-art
methods on a broad range of datasets.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:41:16 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Xu",
"Xin-Jian",
""
],
[
"Deng",
"Chong",
""
],
[
"Zhang",
"Li-Jie",
""
]
] | TITLE: Hyperlink prediction via local random walks and Jensen-Shannon
divergence
ABSTRACT: Many real-world systems involving higher-order interactions can be modeled by
hypergraphs, where vertices represent the systemic units and hyperedges
describe the interactions among them. In this paper, we focus on the problem of
hyperlink prediction which aims at inferring missing hyperlinks based on
observed hyperlinks. We propose three similarity indices for hyperlink
prediction based on local random walks and Jensen-Shannon divergence. Numerical
experiments show that the proposed indices outperform the state-of-the-art
methods on a broad range of datasets.
| no_new_dataset | 0.947769 |
2303.14999 | Zhaofei Wang | Zhaofei Wang, Weijia Zhang, Min-Ling Zhang | Transformer-based Multi-Instance Learning for Weakly Supervised Object
Detection | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Weakly Supervised Object Detection (WSOD) enables the training of object
detection models using only image-level annotations. State-of-the-art WSOD
detectors commonly rely on multi-instance learning (MIL) as the backbone of
their detectors and assume that the bounding box proposals of an image are
independent of each other. However, since such approaches only utilize the
highest score proposal and discard the potentially useful information from
other proposals, their independent MIL backbone often limits models to salient
parts of an object or causes them to detect only one object per class. To solve
the above problems, we propose a novel backbone for WSOD based on our tailored
Vision Transformer named Weakly Supervised Transformer Detection Network
(WSTDN). Our algorithm is not only the first to demonstrate that self-attention
modules that consider inter-instance relationships are effective backbones for
WSOD, but also we introduce a novel bounding box mining method (BBM) integrated
with a memory transfer refinement (MTR) procedure to utilize the instance
dependencies for facilitating instance refinements. Experimental results on
PASCAL VOC2007 and VOC2012 benchmarks demonstrate the effectiveness of our
proposed WSTDN and modified instance refinement modules.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:42:45 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Wang",
"Zhaofei",
""
],
[
"Zhang",
"Weijia",
""
],
[
"Zhang",
"Min-Ling",
""
]
] | TITLE: Transformer-based Multi-Instance Learning for Weakly Supervised Object
Detection
ABSTRACT: Weakly Supervised Object Detection (WSOD) enables the training of object
detection models using only image-level annotations. State-of-the-art WSOD
detectors commonly rely on multi-instance learning (MIL) as the backbone of
their detectors and assume that the bounding box proposals of an image are
independent of each other. However, since such approaches only utilize the
highest score proposal and discard the potentially useful information from
other proposals, their independent MIL backbone often limits models to salient
parts of an object or causes them to detect only one object per class. To solve
the above problems, we propose a novel backbone for WSOD based on our tailored
Vision Transformer named Weakly Supervised Transformer Detection Network
(WSTDN). Our algorithm is not only the first to demonstrate that self-attention
modules that consider inter-instance relationships are effective backbones for
WSOD, but also we introduce a novel bounding box mining method (BBM) integrated
with a memory transfer refinement (MTR) procedure to utilize the instance
dependencies for facilitating instance refinements. Experimental results on
PASCAL VOC2007 and VOC2012 benchmarks demonstrate the effectiveness of our
proposed WSTDN and modified instance refinement modules.
| no_new_dataset | 0.947284 |
2303.15005 | Madalina Erascu | Andreea Postovan and M\u{a}d\u{a}lina Era\c{s}cu | Architecturing Binarized Neural Networks for Traffic Sign Recognition | null | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | Traffic signs support road safety and managing the flow of traffic, hence are
an integral part of any vision system for autonomous driving. While the use of
deep learning is well-known in traffic signs classification due to the high
accuracy results obtained using convolutional neural networks (CNNs) (state of
the art is 99.46\%), little is known about binarized neural networks (BNNs).
Compared to CNNs, BNNs reduce the model size and simplify convolution
operations and have shown promising results in computationally limited and
energy-constrained devices which appear in the context of autonomous driving.
This work presents a bottom-up approach for architecturing BNNs by studying
characteristics of the constituent layers. These constituent layers (binarized
convolutional layers, max pooling, batch normalization, fully connected layers)
are studied in various combinations and with different values of kernel size,
number of filters and of neurons by using the German Traffic Sign Recognition
Benchmark (GTSRB) for training. As a result, we propose BNNs architectures
which achieve more than $90\%$ for GTSRB (the maximum is $96.45\%$) and an
average greater than $80\%$ (the maximum is $88.99\%$) considering also the
Belgian and Chinese datasets for testing. The number of parameters of these
architectures varies from 100k to less than 2M. The accompanying material of
this paper is publicly available at
https://github.com/apostovan21/BinarizedNeuralNetwork.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:46:31 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Postovan",
"Andreea",
""
],
[
"Eraşcu",
"Mădălina",
""
]
] | TITLE: Architecturing Binarized Neural Networks for Traffic Sign Recognition
ABSTRACT: Traffic signs support road safety and managing the flow of traffic, hence are
an integral part of any vision system for autonomous driving. While the use of
deep learning is well-known in traffic signs classification due to the high
accuracy results obtained using convolutional neural networks (CNNs) (state of
the art is 99.46\%), little is known about binarized neural networks (BNNs).
Compared to CNNs, BNNs reduce the model size and simplify convolution
operations and have shown promising results in computationally limited and
energy-constrained devices which appear in the context of autonomous driving.
This work presents a bottom-up approach for architecturing BNNs by studying
characteristics of the constituent layers. These constituent layers (binarized
convolutional layers, max pooling, batch normalization, fully connected layers)
are studied in various combinations and with different values of kernel size,
number of filters and of neurons by using the German Traffic Sign Recognition
Benchmark (GTSRB) for training. As a result, we propose BNNs architectures
which achieve more than $90\%$ for GTSRB (the maximum is $96.45\%$) and an
average greater than $80\%$ (the maximum is $88.99\%$) considering also the
Belgian and Chinese datasets for testing. The number of parameters of these
architectures varies from 100k to less than 2M. The accompanying material of
this paper is publicly available at
https://github.com/apostovan21/BinarizedNeuralNetwork.
| no_new_dataset | 0.954435 |
2303.15006 | Wafa Aissa | Wafa Aissa (CEDRIC - VERTIGO), Marin Ferecatu (CEDRIC - VERTIGO),
Michel Crucianu (CEDRIC - VERTIGO) | Curriculum Learning for Compositional Visual Reasoning | null | In Proceedings of the 18th International Joint Conference on
Computer Vision, Imaging and Computer Graphics Theory and Applications -
Volume 5: VISAPP, Feb 2023, Lisbon, Portugal. pp.888-897 | 10.5220/0011895400003417 | null | cs.CL cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Visual Question Answering (VQA) is a complex task requiring large datasets
and expensive training. Neural Module Networks (NMN) first translate the
question to a reasoning path, then follow that path to analyze the image and
provide an answer. We propose an NMN method that relies on predefined
cross-modal embeddings to ``warm start'' learning on the GQA dataset, then
focus on Curriculum Learning (CL) as a way to improve training and make a
better use of the data. Several difficulty criteria are employed for defining
CL methods. We show that by an appropriate selection of the CL method the cost
of training and the amount of training data can be greatly reduced, with a
limited impact on the final VQA accuracy. Furthermore, we introduce
intermediate losses during training and find that this allows to simplify the
CL strategy.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:47:18 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Aissa",
"Wafa",
"",
"CEDRIC - VERTIGO"
],
[
"Ferecatu",
"Marin",
"",
"CEDRIC - VERTIGO"
],
[
"Crucianu",
"Michel",
"",
"CEDRIC - VERTIGO"
]
] | TITLE: Curriculum Learning for Compositional Visual Reasoning
ABSTRACT: Visual Question Answering (VQA) is a complex task requiring large datasets
and expensive training. Neural Module Networks (NMN) first translate the
question to a reasoning path, then follow that path to analyze the image and
provide an answer. We propose an NMN method that relies on predefined
cross-modal embeddings to ``warm start'' learning on the GQA dataset, then
focus on Curriculum Learning (CL) as a way to improve training and make a
better use of the data. Several difficulty criteria are employed for defining
CL methods. We show that by an appropriate selection of the CL method the cost
of training and the amount of training data can be greatly reduced, with a
limited impact on the final VQA accuracy. Furthermore, we introduce
intermediate losses during training and find that this allows to simplify the
CL strategy.
| no_new_dataset | 0.945551 |
2303.15012 | Senmao Li | Senmao Li, Joost van de Weijer, Yaxing Wang, Fahad Shahbaz Khan,
Meiqin Liu, Jian Yang | 3D-Aware Multi-Class Image-to-Image Translation with NeRFs | Accepted by CVPR2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recent advances in 3D-aware generative models (3D-aware GANs) combined with
Neural Radiance Fields (NeRF) have achieved impressive results. However no
prior works investigate 3D-aware GANs for 3D consistent multi-class
image-to-image (3D-aware I2I) translation. Naively using 2D-I2I translation
methods suffers from unrealistic shape/identity change. To perform 3D-aware
multi-class I2I translation, we decouple this learning process into a
multi-class 3D-aware GAN step and a 3D-aware I2I translation step. In the first
step, we propose two novel techniques: a new conditional architecture and an
effective training strategy. In the second step, based on the well-trained
multi-class 3D-aware GAN architecture, that preserves view-consistency, we
construct a 3D-aware I2I translation system. To further reduce the
view-consistency problems, we propose several new techniques, including a
U-net-like adaptor network design, a hierarchical representation constrain and
a relative regularization loss. In extensive experiments on two datasets,
quantitative and qualitative results demonstrate that we successfully perform
3D-aware I2I translation with multi-view consistency.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:54:51 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Li",
"Senmao",
""
],
[
"van de Weijer",
"Joost",
""
],
[
"Wang",
"Yaxing",
""
],
[
"Khan",
"Fahad Shahbaz",
""
],
[
"Liu",
"Meiqin",
""
],
[
"Yang",
"Jian",
""
]
] | TITLE: 3D-Aware Multi-Class Image-to-Image Translation with NeRFs
ABSTRACT: Recent advances in 3D-aware generative models (3D-aware GANs) combined with
Neural Radiance Fields (NeRF) have achieved impressive results. However no
prior works investigate 3D-aware GANs for 3D consistent multi-class
image-to-image (3D-aware I2I) translation. Naively using 2D-I2I translation
methods suffers from unrealistic shape/identity change. To perform 3D-aware
multi-class I2I translation, we decouple this learning process into a
multi-class 3D-aware GAN step and a 3D-aware I2I translation step. In the first
step, we propose two novel techniques: a new conditional architecture and an
effective training strategy. In the second step, based on the well-trained
multi-class 3D-aware GAN architecture, that preserves view-consistency, we
construct a 3D-aware I2I translation system. To further reduce the
view-consistency problems, we propose several new techniques, including a
U-net-like adaptor network design, a hierarchical representation constrain and
a relative regularization loss. In extensive experiments on two datasets,
quantitative and qualitative results demonstrate that we successfully perform
3D-aware I2I translation with multi-view consistency.
| no_new_dataset | 0.943764 |
2303.15014 | Hyun Seok Seong | Hyun Seok Seong, WonJun Moon, SuBeen Lee, Jae-Pil Heo | Leveraging Hidden Positives for Unsupervised Semantic Segmentation | Accepted to CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Dramatic demand for manpower to label pixel-level annotations triggered the
advent of unsupervised semantic segmentation. Although the recent work
employing the vision transformer (ViT) backbone shows exceptional performance,
there is still a lack of consideration for task-specific training guidance and
local semantic consistency. To tackle these issues, we leverage contrastive
learning by excavating hidden positives to learn rich semantic relationships
and ensure semantic consistency in local regions. Specifically, we first
discover two types of global hidden positives, task-agnostic and task-specific
ones for each anchor based on the feature similarities defined by a fixed
pre-trained backbone and a segmentation head-in-training, respectively. A
gradual increase in the contribution of the latter induces the model to capture
task-specific semantic features. In addition, we introduce a gradient
propagation strategy to learn semantic consistency between adjacent patches,
under the inherent premise that nearby patches are highly likely to possess the
same semantics. Specifically, we add the loss propagating to local hidden
positives, semantically similar nearby patches, in proportion to the predefined
similarity scores. With these training schemes, our proposed method achieves
new state-of-the-art (SOTA) results in COCO-stuff, Cityscapes, and Potsdam-3
datasets. Our code is available at: https://github.com/hynnsk/HP.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:57:28 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Seong",
"Hyun Seok",
""
],
[
"Moon",
"WonJun",
""
],
[
"Lee",
"SuBeen",
""
],
[
"Heo",
"Jae-Pil",
""
]
] | TITLE: Leveraging Hidden Positives for Unsupervised Semantic Segmentation
ABSTRACT: Dramatic demand for manpower to label pixel-level annotations triggered the
advent of unsupervised semantic segmentation. Although the recent work
employing the vision transformer (ViT) backbone shows exceptional performance,
there is still a lack of consideration for task-specific training guidance and
local semantic consistency. To tackle these issues, we leverage contrastive
learning by excavating hidden positives to learn rich semantic relationships
and ensure semantic consistency in local regions. Specifically, we first
discover two types of global hidden positives, task-agnostic and task-specific
ones for each anchor based on the feature similarities defined by a fixed
pre-trained backbone and a segmentation head-in-training, respectively. A
gradual increase in the contribution of the latter induces the model to capture
task-specific semantic features. In addition, we introduce a gradient
propagation strategy to learn semantic consistency between adjacent patches,
under the inherent premise that nearby patches are highly likely to possess the
same semantics. Specifically, we add the loss propagating to local hidden
positives, semantically similar nearby patches, in proportion to the predefined
similarity scores. With these training schemes, our proposed method achieves
new state-of-the-art (SOTA) results in COCO-stuff, Cityscapes, and Potsdam-3
datasets. Our code is available at: https://github.com/hynnsk/HP.
| no_new_dataset | 0.944944 |
2303.15016 | Chunpu Xu | Chunpu Xu and Jing Li | Borrowing Human Senses: Comment-Aware Self-Training for Social Media
Multimodal Classification | accepted to EMNLP 2022 | null | null | null | cs.CL cs.AI cs.IR cs.MM | http://creativecommons.org/licenses/by/4.0/ | Social media is daily creating massive multimedia content with paired image
and text, presenting the pressing need to automate the vision and language
understanding for various multimodal classification tasks. Compared to the
commonly researched visual-lingual data, social media posts tend to exhibit
more implicit image-text relations. To better glue the cross-modal semantics
therein, we capture hinting features from user comments, which are retrieved
via jointly leveraging visual and lingual similarity. Afterwards, the
classification tasks are explored via self-training in a teacher-student
framework, motivated by the usually limited labeled data scales in existing
benchmarks. Substantial experiments are conducted on four multimodal social
media benchmarks for image text relation classification, sarcasm detection,
sentiment classification, and hate speech detection. The results show that our
method further advances the performance of previous state-of-the-art models,
which do not employ comment modeling or self-training.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 08:59:55 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Xu",
"Chunpu",
""
],
[
"Li",
"Jing",
""
]
] | TITLE: Borrowing Human Senses: Comment-Aware Self-Training for Social Media
Multimodal Classification
ABSTRACT: Social media is daily creating massive multimedia content with paired image
and text, presenting the pressing need to automate the vision and language
understanding for various multimodal classification tasks. Compared to the
commonly researched visual-lingual data, social media posts tend to exhibit
more implicit image-text relations. To better glue the cross-modal semantics
therein, we capture hinting features from user comments, which are retrieved
via jointly leveraging visual and lingual similarity. Afterwards, the
classification tasks are explored via self-training in a teacher-student
framework, motivated by the usually limited labeled data scales in existing
benchmarks. Substantial experiments are conducted on four multimodal social
media benchmarks for image text relation classification, sarcasm detection,
sentiment classification, and hate speech detection. The results show that our
method further advances the performance of previous state-of-the-art models,
which do not employ comment modeling or self-training.
| no_new_dataset | 0.947962 |
2303.15023 | Chen Li | Chen Li and Gim Hee Lee | ScarceNet: Animal Pose Estimation with Scarce Annotations | Accepted to CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Animal pose estimation is an important but under-explored task due to the
lack of labeled data. In this paper, we tackle the task of animal pose
estimation with scarce annotations, where only a small set of labeled data and
unlabeled images are available. At the core of the solution to this problem
setting is the use of the unlabeled data to compensate for the lack of
well-labeled animal pose data. To this end, we propose the ScarceNet, a pseudo
label-based approach to generate artificial labels for the unlabeled images.
The pseudo labels, which are generated with a model trained with the small set
of labeled images, are generally noisy and can hurt the performance when
directly used for training. To solve this problem, we first use a small-loss
trick to select reliable pseudo labels. Although effective, the selection
process is improvident since numerous high-loss samples are left unused. We
further propose to identify reusable samples from the high-loss samples based
on an agreement check. Pseudo labels are re-generated to provide supervision
for those reusable samples. Lastly, we introduce a student-teacher framework to
enforce a consistency constraint since there are still samples that are neither
reliable nor reusable. By combining the reliable pseudo label selection with
the reusable sample re-labeling and the consistency constraint, we can make
full use of the unlabeled data. We evaluate our approach on the challenging
AP-10K dataset, where our approach outperforms existing semi-supervised
approaches by a large margin. We also test on the TigDog dataset, where our
approach can achieve better performance than domain adaptation based approaches
when only very few annotations are available. Our code is available at the
project website.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 09:15:53 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Li",
"Chen",
""
],
[
"Lee",
"Gim Hee",
""
]
] | TITLE: ScarceNet: Animal Pose Estimation with Scarce Annotations
ABSTRACT: Animal pose estimation is an important but under-explored task due to the
lack of labeled data. In this paper, we tackle the task of animal pose
estimation with scarce annotations, where only a small set of labeled data and
unlabeled images are available. At the core of the solution to this problem
setting is the use of the unlabeled data to compensate for the lack of
well-labeled animal pose data. To this end, we propose the ScarceNet, a pseudo
label-based approach to generate artificial labels for the unlabeled images.
The pseudo labels, which are generated with a model trained with the small set
of labeled images, are generally noisy and can hurt the performance when
directly used for training. To solve this problem, we first use a small-loss
trick to select reliable pseudo labels. Although effective, the selection
process is improvident since numerous high-loss samples are left unused. We
further propose to identify reusable samples from the high-loss samples based
on an agreement check. Pseudo labels are re-generated to provide supervision
for those reusable samples. Lastly, we introduce a student-teacher framework to
enforce a consistency constraint since there are still samples that are neither
reliable nor reusable. By combining the reliable pseudo label selection with
the reusable sample re-labeling and the consistency constraint, we can make
full use of the unlabeled data. We evaluate our approach on the challenging
AP-10K dataset, where our approach outperforms existing semi-supervised
approaches by a large margin. We also test on the TigDog dataset, where our
approach can achieve better performance than domain adaptation based approaches
when only very few annotations are available. Our code is available at the
project website.
| no_new_dataset | 0.946745 |
2303.15024 | Xiaoyu Bai | Xiaoyu Bai, Yong Xia | An End-to-End Framework For Universal Lesion Detection With Missing
Annotations | null | null | 10.1109/ICSP56322.2022.9965335 | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Fully annotated large-scale medical image datasets are highly valuable.
However, because labeling medical images is tedious and requires specialized
knowledge, the large-scale datasets available often have missing annotation
issues. For instance, DeepLesion, a large-scale CT image dataset with labels
for various kinds of lesions, is reported to have a missing annotation rate of
50\%. Directly training a lesion detector on it would suffer from false
negative supervision caused by unannotated lesions. To address this issue,
previous works have used sophisticated multi-stage strategies to switch between
lesion mining and detector training. In this work, we present a novel
end-to-end framework for mining unlabeled lesions while simultaneously training
the detector. Our framework follows the teacher-student paradigm. In each
iteration, the teacher model infers the input data and creates a set of
predictions. High-confidence predictions are combined with partially-labeled
ground truth for training the student model. On the DeepLesion dataset, using
the original partially labeled training set, our model can outperform all other
more complicated methods and surpass the previous best method by 2.3\% on
average sensitivity and 2.7\% on average precision, achieving state-of-the-art
universal lesion detection results.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 09:16:10 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Bai",
"Xiaoyu",
""
],
[
"Xia",
"Yong",
""
]
] | TITLE: An End-to-End Framework For Universal Lesion Detection With Missing
Annotations
ABSTRACT: Fully annotated large-scale medical image datasets are highly valuable.
However, because labeling medical images is tedious and requires specialized
knowledge, the large-scale datasets available often have missing annotation
issues. For instance, DeepLesion, a large-scale CT image dataset with labels
for various kinds of lesions, is reported to have a missing annotation rate of
50\%. Directly training a lesion detector on it would suffer from false
negative supervision caused by unannotated lesions. To address this issue,
previous works have used sophisticated multi-stage strategies to switch between
lesion mining and detector training. In this work, we present a novel
end-to-end framework for mining unlabeled lesions while simultaneously training
the detector. Our framework follows the teacher-student paradigm. In each
iteration, the teacher model infers the input data and creates a set of
predictions. High-confidence predictions are combined with partially-labeled
ground truth for training the student model. On the DeepLesion dataset, using
the original partially labeled training set, our model can outperform all other
more complicated methods and surpass the previous best method by 2.3\% on
average sensitivity and 2.7\% on average precision, achieving state-of-the-art
universal lesion detection results.
| no_new_dataset | 0.877529 |
2303.15043 | Wei Shang | Wei Shang, Dongwei Ren, Yi Yang, Hongzhi Zhang, Kede Ma, Wangmeng Zuo | Joint Video Multi-Frame Interpolation and Deblurring under Unknown
Exposure Time | Accepted by CVPR 2023, available at
https://github.com/shangwei5/VIDUE | null | null | null | cs.CV eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Natural videos captured by consumer cameras often suffer from low framerate
and motion blur due to the combination of dynamic scene complexity, lens and
sensor imperfection, and less than ideal exposure setting. As a result,
computational methods that jointly perform video frame interpolation and
deblurring begin to emerge with the unrealistic assumption that the exposure
time is known and fixed. In this work, we aim ambitiously for a more realistic
and challenging task - joint video multi-frame interpolation and deblurring
under unknown exposure time. Toward this goal, we first adopt a variant of
supervised contrastive learning to construct an exposure-aware representation
from input blurred frames. We then train two U-Nets for intra-motion and
inter-motion analysis, respectively, adapting to the learned exposure
representation via gain tuning. We finally build our video reconstruction
network upon the exposure and motion representation by progressive
exposure-adaptive convolution and motion refinement. Extensive experiments on
both simulated and real-world datasets show that our optimized method achieves
notable performance gains over the state-of-the-art on the joint video x8
interpolation and deblurring task. Moreover, on the seemingly implausible x16
interpolation task, our method outperforms existing methods by more than 1.5 dB
in terms of PSNR.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 09:43:42 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Shang",
"Wei",
""
],
[
"Ren",
"Dongwei",
""
],
[
"Yang",
"Yi",
""
],
[
"Zhang",
"Hongzhi",
""
],
[
"Ma",
"Kede",
""
],
[
"Zuo",
"Wangmeng",
""
]
] | TITLE: Joint Video Multi-Frame Interpolation and Deblurring under Unknown
Exposure Time
ABSTRACT: Natural videos captured by consumer cameras often suffer from low framerate
and motion blur due to the combination of dynamic scene complexity, lens and
sensor imperfection, and less than ideal exposure setting. As a result,
computational methods that jointly perform video frame interpolation and
deblurring begin to emerge with the unrealistic assumption that the exposure
time is known and fixed. In this work, we aim ambitiously for a more realistic
and challenging task - joint video multi-frame interpolation and deblurring
under unknown exposure time. Toward this goal, we first adopt a variant of
supervised contrastive learning to construct an exposure-aware representation
from input blurred frames. We then train two U-Nets for intra-motion and
inter-motion analysis, respectively, adapting to the learned exposure
representation via gain tuning. We finally build our video reconstruction
network upon the exposure and motion representation by progressive
exposure-adaptive convolution and motion refinement. Extensive experiments on
both simulated and real-world datasets show that our optimized method achieves
notable performance gains over the state-of-the-art on the joint video x8
interpolation and deblurring task. Moreover, on the seemingly implausible x16
interpolation task, our method outperforms existing methods by more than 1.5 dB
in terms of PSNR.
| no_new_dataset | 0.943815 |
2303.15046 | Yuekun Dai | Yuekun Dai, Yihang Luo, Shangchen Zhou, Chongyi Li, Chen Change Loy | Nighttime Smartphone Reflective Flare Removal Using Optical Center
Symmetry Prior | CVPR2023 (Highlight) | null | null | null | cs.CV eess.IV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Reflective flare is a phenomenon that occurs when light reflects inside
lenses, causing bright spots or a "ghosting effect" in photos, which can impact
their quality. Eliminating reflective flare is highly desirable but
challenging. Many existing methods rely on manually designed features to detect
these bright spots, but they often fail to identify reflective flares created
by various types of light and may even mistakenly remove the light sources in
scenarios with multiple light sources. To address these challenges, we propose
an optical center symmetry prior, which suggests that the reflective flare and
light source are always symmetrical around the lens's optical center. This
prior helps to locate the reflective flare's proposal region more accurately
and can be applied to most smartphone cameras. Building on this prior, we
create the first reflective flare removal dataset called BracketFlare, which
contains diverse and realistic reflective flare patterns. We use continuous
bracketing to capture the reflective flare pattern in the underexposed image
and combine it with a normally exposed image to synthesize a pair of
flare-corrupted and flare-free images. With the dataset, neural networks can be
trained to remove the reflective flares effectively. Extensive experiments
demonstrate the effectiveness of our method on both synthetic and real-world
datasets.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 09:44:40 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Dai",
"Yuekun",
""
],
[
"Luo",
"Yihang",
""
],
[
"Zhou",
"Shangchen",
""
],
[
"Li",
"Chongyi",
""
],
[
"Loy",
"Chen Change",
""
]
] | TITLE: Nighttime Smartphone Reflective Flare Removal Using Optical Center
Symmetry Prior
ABSTRACT: Reflective flare is a phenomenon that occurs when light reflects inside
lenses, causing bright spots or a "ghosting effect" in photos, which can impact
their quality. Eliminating reflective flare is highly desirable but
challenging. Many existing methods rely on manually designed features to detect
these bright spots, but they often fail to identify reflective flares created
by various types of light and may even mistakenly remove the light sources in
scenarios with multiple light sources. To address these challenges, we propose
an optical center symmetry prior, which suggests that the reflective flare and
light source are always symmetrical around the lens's optical center. This
prior helps to locate the reflective flare's proposal region more accurately
and can be applied to most smartphone cameras. Building on this prior, we
create the first reflective flare removal dataset called BracketFlare, which
contains diverse and realistic reflective flare patterns. We use continuous
bracketing to capture the reflective flare pattern in the underexposed image
and combine it with a normally exposed image to synthesize a pair of
flare-corrupted and flare-free images. With the dataset, neural networks can be
trained to remove the reflective flares effectively. Extensive experiments
demonstrate the effectiveness of our method on both synthetic and real-world
datasets.
| new_dataset | 0.961822 |
2303.15053 | Eric Gustaf Ted Wulff | Eric Wulff, Maria Girone, David Southwick, Juan Pablo Garc\'ia
Amboage, Eduard Cuba | Hyperparameter optimization, quantum-assisted model performance
prediction, and benchmarking of AI-based High Energy Physics workloads using
HPC | 5 pages, 7 figures. Submitted to the proceedings of the ACAT 2022
conference and is to be published in the Journal Of Physics: Conference
Series | null | null | null | physics.data-an cs.LG | http://creativecommons.org/licenses/by/4.0/ | Training and Hyperparameter Optimization (HPO) of deep learning-based AI
models are often compute resource intensive and calls for the use of
large-scale distributed resources as well as scalable and resource efficient
hyperparameter search algorithms. This work studies the potential of using
model performance prediction to aid the HPO process carried out on High
Performance Computing systems. In addition, a quantum annealer is used to train
the performance predictor and a method is proposed to overcome some of the
problems derived from the current limitations in quantum systems as well as to
increase the stability of solutions. This allows for achieving results on a
quantum machine comparable to those obtained on a classical machine, showing
how quantum computers could be integrated within classical machine learning
tuning pipelines.
Furthermore, results are presented from the development of a containerized
benchmark based on an AI-model for collision event reconstruction that allows
us to compare and assess the suitability of different hardware accelerators for
training deep neural networks.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 09:55:33 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Wulff",
"Eric",
""
],
[
"Girone",
"Maria",
""
],
[
"Southwick",
"David",
""
],
[
"Amboage",
"Juan Pablo García",
""
],
[
"Cuba",
"Eduard",
""
]
] | TITLE: Hyperparameter optimization, quantum-assisted model performance
prediction, and benchmarking of AI-based High Energy Physics workloads using
HPC
ABSTRACT: Training and Hyperparameter Optimization (HPO) of deep learning-based AI
models are often compute resource intensive and calls for the use of
large-scale distributed resources as well as scalable and resource efficient
hyperparameter search algorithms. This work studies the potential of using
model performance prediction to aid the HPO process carried out on High
Performance Computing systems. In addition, a quantum annealer is used to train
the performance predictor and a method is proposed to overcome some of the
problems derived from the current limitations in quantum systems as well as to
increase the stability of solutions. This allows for achieving results on a
quantum machine comparable to those obtained on a classical machine, showing
how quantum computers could be integrated within classical machine learning
tuning pipelines.
Furthermore, results are presented from the development of a containerized
benchmark based on an AI-model for collision event reconstruction that allows
us to compare and assess the suitability of different hardware accelerators for
training deep neural networks.
| no_new_dataset | 0.942082 |
2303.15062 | Beomyoung Kim | Beomyoung Kim, Joonhyun Jeong, Dongyoon Han, Sung Ju Hwang | The Devil is in the Points: Weakly Semi-Supervised Instance Segmentation
via Point-Guided Mask Representation | CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we introduce a novel learning scheme named weakly
semi-supervised instance segmentation (WSSIS) with point labels for
budget-efficient and high-performance instance segmentation. Namely, we
consider a dataset setting consisting of a few fully-labeled images and a lot
of point-labeled images. Motivated by the main challenge of semi-supervised
approaches mainly derives from the trade-off between false-negative and
false-positive instance proposals, we propose a method for WSSIS that can
effectively leverage the budget-friendly point labels as a powerful weak
supervision source to resolve the challenge. Furthermore, to deal with the hard
case where the amount of fully-labeled data is extremely limited, we propose a
MaskRefineNet that refines noise in rough masks. We conduct extensive
experiments on COCO and BDD100K datasets, and the proposed method achieves
promising results comparable to those of the fully-supervised model, even with
50% of the fully labeled COCO data (38.8% vs. 39.7%). Moreover, when using as
little as 5% of fully labeled COCO data, our method shows significantly
superior performance over the state-of-the-art semi-supervised learning method
(33.7% vs. 24.9%). The code is available at
https://github.com/clovaai/PointWSSIS.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 10:11:22 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Kim",
"Beomyoung",
""
],
[
"Jeong",
"Joonhyun",
""
],
[
"Han",
"Dongyoon",
""
],
[
"Hwang",
"Sung Ju",
""
]
] | TITLE: The Devil is in the Points: Weakly Semi-Supervised Instance Segmentation
via Point-Guided Mask Representation
ABSTRACT: In this paper, we introduce a novel learning scheme named weakly
semi-supervised instance segmentation (WSSIS) with point labels for
budget-efficient and high-performance instance segmentation. Namely, we
consider a dataset setting consisting of a few fully-labeled images and a lot
of point-labeled images. Motivated by the main challenge of semi-supervised
approaches mainly derives from the trade-off between false-negative and
false-positive instance proposals, we propose a method for WSSIS that can
effectively leverage the budget-friendly point labels as a powerful weak
supervision source to resolve the challenge. Furthermore, to deal with the hard
case where the amount of fully-labeled data is extremely limited, we propose a
MaskRefineNet that refines noise in rough masks. We conduct extensive
experiments on COCO and BDD100K datasets, and the proposed method achieves
promising results comparable to those of the fully-supervised model, even with
50% of the fully labeled COCO data (38.8% vs. 39.7%). Moreover, when using as
little as 5% of fully labeled COCO data, our method shows significantly
superior performance over the state-of-the-art semi-supervised learning method
(33.7% vs. 24.9%). The code is available at
https://github.com/clovaai/PointWSSIS.
| no_new_dataset | 0.951414 |
2303.15068 | Bestoun Ahmed Dr. | Firas Bayram, Bestoun S. Ahmed, Erik Hallin, Anton Engman | DQSOps: Data Quality Scoring Operations Framework for Data-Driven
Applications | 10 Pages The International Conference on Evaluation and Assessment in
Software Engineering (EASE) conference | null | null | null | cs.DB cs.SE | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Data quality assessment has become a prominent component in the successful
execution of complex data-driven artificial intelligence (AI) software systems.
In practice, real-world applications generate huge volumes of data at speeds.
These data streams require analysis and preprocessing before being permanently
stored or used in a learning task. Therefore, significant attention has been
paid to the systematic management and construction of high-quality datasets.
Nevertheless, managing voluminous and high-velocity data streams is usually
performed manually (i.e. offline), making it an impractical strategy in
production environments. To address this challenge, DataOps has emerged to
achieve life-cycle automation of data processes using DevOps principles.
However, determining the data quality based on a fitness scale constitutes a
complex task within the framework of DataOps. This paper presents a novel Data
Quality Scoring Operations (DQSOps) framework that yields a quality score for
production data in DataOps workflows. The framework incorporates two scoring
approaches, an ML prediction-based approach that predicts the data quality
score and a standard-based approach that periodically produces the ground-truth
scores based on assessing several data quality dimensions. We deploy the DQSOps
framework in a real-world industrial use case. The results show that DQSOps
achieves significant computational speedup rates compared to the conventional
approach of data quality scoring while maintaining high prediction performance.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 10:20:38 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Bayram",
"Firas",
""
],
[
"Ahmed",
"Bestoun S.",
""
],
[
"Hallin",
"Erik",
""
],
[
"Engman",
"Anton",
""
]
] | TITLE: DQSOps: Data Quality Scoring Operations Framework for Data-Driven
Applications
ABSTRACT: Data quality assessment has become a prominent component in the successful
execution of complex data-driven artificial intelligence (AI) software systems.
In practice, real-world applications generate huge volumes of data at speeds.
These data streams require analysis and preprocessing before being permanently
stored or used in a learning task. Therefore, significant attention has been
paid to the systematic management and construction of high-quality datasets.
Nevertheless, managing voluminous and high-velocity data streams is usually
performed manually (i.e. offline), making it an impractical strategy in
production environments. To address this challenge, DataOps has emerged to
achieve life-cycle automation of data processes using DevOps principles.
However, determining the data quality based on a fitness scale constitutes a
complex task within the framework of DataOps. This paper presents a novel Data
Quality Scoring Operations (DQSOps) framework that yields a quality score for
production data in DataOps workflows. The framework incorporates two scoring
approaches, an ML prediction-based approach that predicts the data quality
score and a standard-based approach that periodically produces the ground-truth
scores based on assessing several data quality dimensions. We deploy the DQSOps
framework in a real-world industrial use case. The results show that DQSOps
achieves significant computational speedup rates compared to the conventional
approach of data quality scoring while maintaining high prediction performance.
| no_new_dataset | 0.949106 |
2303.15097 | Markus G\"olles | Sandra Staudt, Viktor Unterberger, Markus G\"olles, Michael Wernhart,
Ren\'e Rieberer, Martin Horn | Control-oriented modeling of a LiBr/H2O absorption heat pumping device
and experimental validation | null | null | null | null | eess.SY cs.SY | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Absorption heat pumping devices (AHPDs, comprising absorption heat pumps and
chillers) are devices that use thermal energy instead of electricity to
generate heating and cooling, thereby facilitating the use of waste heat and
renewable energy sources such as solar or geothermal energy. Despite this
benefit, widespread use of AHPDs is still limited. One reason for this is
partly unsatisfactory control performance under varying operating conditions,
which can result in poor modulation and part load capability. A promising
approach to tackle this issue is using dynamic, model-based control strategies,
whose effectiveness, however, strongly depend on the model being used. This
paper therefore focuses on the derivation of a viable dynamic model to be used
for such model-based control strategies for AHPDs such as state feedback or
model-predictive control. The derived model is experimentally validated,
showing good modeling accuracy. Its modeling accuracy is also compared to
alternative model versions, that contain other heat transfer correlations, as a
benchmark. Although the derived model is mathematically simple, it does have
the structure of a nonlinear differential-algebraic system of equations. To
obtain an even simpler model structure, linearization at an operating point is
discussed to derive a model in linear state space representation. The
experimental validation shows that the linear model does have slightly worse
steady-state accuracy, but that the dynamic accuracy seems to be almost
unaffected by the linearization. The presented new modeling approach is
considered suitable to be used as a basis for the design of advanced,
model-based control strategies, ultimately aiming to improve the modulation and
part load capability of AHPDs.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:03:20 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Staudt",
"Sandra",
""
],
[
"Unterberger",
"Viktor",
""
],
[
"Gölles",
"Markus",
""
],
[
"Wernhart",
"Michael",
""
],
[
"Rieberer",
"René",
""
],
[
"Horn",
"Martin",
""
]
] | TITLE: Control-oriented modeling of a LiBr/H2O absorption heat pumping device
and experimental validation
ABSTRACT: Absorption heat pumping devices (AHPDs, comprising absorption heat pumps and
chillers) are devices that use thermal energy instead of electricity to
generate heating and cooling, thereby facilitating the use of waste heat and
renewable energy sources such as solar or geothermal energy. Despite this
benefit, widespread use of AHPDs is still limited. One reason for this is
partly unsatisfactory control performance under varying operating conditions,
which can result in poor modulation and part load capability. A promising
approach to tackle this issue is using dynamic, model-based control strategies,
whose effectiveness, however, strongly depend on the model being used. This
paper therefore focuses on the derivation of a viable dynamic model to be used
for such model-based control strategies for AHPDs such as state feedback or
model-predictive control. The derived model is experimentally validated,
showing good modeling accuracy. Its modeling accuracy is also compared to
alternative model versions, that contain other heat transfer correlations, as a
benchmark. Although the derived model is mathematically simple, it does have
the structure of a nonlinear differential-algebraic system of equations. To
obtain an even simpler model structure, linearization at an operating point is
discussed to derive a model in linear state space representation. The
experimental validation shows that the linear model does have slightly worse
steady-state accuracy, but that the dynamic accuracy seems to be almost
unaffected by the linearization. The presented new modeling approach is
considered suitable to be used as a basis for the design of advanced,
model-based control strategies, ultimately aiming to improve the modulation and
part load capability of AHPDs.
| no_new_dataset | 0.942295 |
2303.15105 | Qiming Zhang | Qiming Zhang, Jing Zhang, Yufei Xu, Dacheng Tao | Vision Transformer with Quadrangle Attention | 15 pages, the extension of the ECCV 2022 paper (VSA: Learning
Varied-Size Window Attention in Vision Transformers) | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Window-based attention has become a popular choice in vision transformers due
to its superior performance, lower computational complexity, and less memory
footprint. However, the design of hand-crafted windows, which is data-agnostic,
constrains the flexibility of transformers to adapt to objects of varying
sizes, shapes, and orientations. To address this issue, we propose a novel
quadrangle attention (QA) method that extends the window-based attention to a
general quadrangle formulation. Our method employs an end-to-end learnable
quadrangle regression module that predicts a transformation matrix to transform
default windows into target quadrangles for token sampling and attention
calculation, enabling the network to model various targets with different
shapes and orientations and capture rich context information. We integrate QA
into plain and hierarchical vision transformers to create a new architecture
named QFormer, which offers minor code modifications and negligible extra
computational cost. Extensive experiments on public benchmarks demonstrate that
QFormer outperforms existing representative vision transformers on various
vision tasks, including classification, object detection, semantic
segmentation, and pose estimation. The code will be made publicly available at
\href{https://github.com/ViTAE-Transformer/QFormer}{QFormer}.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:13:50 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Zhang",
"Qiming",
""
],
[
"Zhang",
"Jing",
""
],
[
"Xu",
"Yufei",
""
],
[
"Tao",
"Dacheng",
""
]
] | TITLE: Vision Transformer with Quadrangle Attention
ABSTRACT: Window-based attention has become a popular choice in vision transformers due
to its superior performance, lower computational complexity, and less memory
footprint. However, the design of hand-crafted windows, which is data-agnostic,
constrains the flexibility of transformers to adapt to objects of varying
sizes, shapes, and orientations. To address this issue, we propose a novel
quadrangle attention (QA) method that extends the window-based attention to a
general quadrangle formulation. Our method employs an end-to-end learnable
quadrangle regression module that predicts a transformation matrix to transform
default windows into target quadrangles for token sampling and attention
calculation, enabling the network to model various targets with different
shapes and orientations and capture rich context information. We integrate QA
into plain and hierarchical vision transformers to create a new architecture
named QFormer, which offers minor code modifications and negligible extra
computational cost. Extensive experiments on public benchmarks demonstrate that
QFormer outperforms existing representative vision transformers on various
vision tasks, including classification, object detection, semantic
segmentation, and pose estimation. The code will be made publicly available at
\href{https://github.com/ViTAE-Transformer/QFormer}{QFormer}.
| no_new_dataset | 0.950411 |
2303.15107 | Baichun Wei | Baichun Wei, Chunzhi Yi, Qi Zhang, Haiqi Zhu, Jianfei Zhu and Feng
Jiang | ActiveSelfHAR: Incorporating Self Training into Active Learning to
Improve Cross-Subject Human Activity Recognition | null | null | null | null | cs.HC | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Deep learning-based human activity recognition (HAR) methods have shown great
promise in the applications of smart healthcare systems and wireless body
sensor network (BSN). Despite their demonstrated performance in laboratory
settings, the real-world implementation of such methods is still hindered by
the cross-subject issue when adapting to new users. To solve this issue, we
propose ActiveSelfHAR, a framework that combines active learning's benefit of
sparsely acquiring data with actual labels and self- training's benefit of
effectively utilizing unlabeled data to enable the deep model to adapt to the
target domain, i.e., the new users. In this framework, the model trained in the
last iteration or the source domain is first utilized to generate pseudo labels
of the target-domain samples and construct a self-training set based on the
confidence score. Second, we propose to use the spatio-temporal relationships
among the samples in the non-self-training set to augment the core set selected
by active learning. Finally, we combine the self-training set and the augmented
core set to fine-tune the model. We demonstrate our method by comparing it with
state-of-the-art methods on two IMU-based datasets and an EMG-based dataset.
Our method presents similar HAR accuracies with the upper bound, i.e. fully
supervised fine-tuning with less than 1\% labeled data of the target dataset
and significantly improves data efficiency and time cost. Our work highlights
the potential of implementing user-independent HAR methods into smart
healthcare systems and BSN.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:22:08 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Wei",
"Baichun",
""
],
[
"Yi",
"Chunzhi",
""
],
[
"Zhang",
"Qi",
""
],
[
"Zhu",
"Haiqi",
""
],
[
"Zhu",
"Jianfei",
""
],
[
"Jiang",
"Feng",
""
]
] | TITLE: ActiveSelfHAR: Incorporating Self Training into Active Learning to
Improve Cross-Subject Human Activity Recognition
ABSTRACT: Deep learning-based human activity recognition (HAR) methods have shown great
promise in the applications of smart healthcare systems and wireless body
sensor network (BSN). Despite their demonstrated performance in laboratory
settings, the real-world implementation of such methods is still hindered by
the cross-subject issue when adapting to new users. To solve this issue, we
propose ActiveSelfHAR, a framework that combines active learning's benefit of
sparsely acquiring data with actual labels and self- training's benefit of
effectively utilizing unlabeled data to enable the deep model to adapt to the
target domain, i.e., the new users. In this framework, the model trained in the
last iteration or the source domain is first utilized to generate pseudo labels
of the target-domain samples and construct a self-training set based on the
confidence score. Second, we propose to use the spatio-temporal relationships
among the samples in the non-self-training set to augment the core set selected
by active learning. Finally, we combine the self-training set and the augmented
core set to fine-tune the model. We demonstrate our method by comparing it with
state-of-the-art methods on two IMU-based datasets and an EMG-based dataset.
Our method presents similar HAR accuracies with the upper bound, i.e. fully
supervised fine-tuning with less than 1\% labeled data of the target dataset
and significantly improves data efficiency and time cost. Our work highlights
the potential of implementing user-independent HAR methods into smart
healthcare systems and BSN.
| no_new_dataset | 0.948058 |
2303.15108 | Sang-Hun Han | Sang-Hun Han, Min-Gyu Park, Ju Hong Yoon, Ju-Mi Kang, Young-Jae Park
and Hae-Gon Jeon | High-fidelity 3D Human Digitization from Single 2K Resolution Images | code page : https://github.com/SangHunHan92/2K2K, Accepted to CVPR
2023 (Highlight) | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | High-quality 3D human body reconstruction requires high-fidelity and
large-scale training data and appropriate network design that effectively
exploits the high-resolution input images. To tackle these problems, we propose
a simple yet effective 3D human digitization method called 2K2K, which
constructs a large-scale 2K human dataset and infers 3D human models from 2K
resolution images. The proposed method separately recovers the global shape of
a human and its details. The low-resolution depth network predicts the global
structure from a low-resolution image, and the part-wise image-to-normal
network predicts the details of the 3D human body structure. The
high-resolution depth network merges the global 3D shape and the detailed
structures to infer the high-resolution front and back side depth maps.
Finally, an off-the-shelf mesh generator reconstructs the full 3D human model,
which are available at https://github.com/SangHunHan92/2K2K. In addition, we
also provide 2,050 3D human models, including texture maps, 3D joints, and SMPL
parameters for research purposes. In experiments, we demonstrate competitive
performance over the recent works on various datasets.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:22:54 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Han",
"Sang-Hun",
""
],
[
"Park",
"Min-Gyu",
""
],
[
"Yoon",
"Ju Hong",
""
],
[
"Kang",
"Ju-Mi",
""
],
[
"Park",
"Young-Jae",
""
],
[
"Jeon",
"Hae-Gon",
""
]
] | TITLE: High-fidelity 3D Human Digitization from Single 2K Resolution Images
ABSTRACT: High-quality 3D human body reconstruction requires high-fidelity and
large-scale training data and appropriate network design that effectively
exploits the high-resolution input images. To tackle these problems, we propose
a simple yet effective 3D human digitization method called 2K2K, which
constructs a large-scale 2K human dataset and infers 3D human models from 2K
resolution images. The proposed method separately recovers the global shape of
a human and its details. The low-resolution depth network predicts the global
structure from a low-resolution image, and the part-wise image-to-normal
network predicts the details of the 3D human body structure. The
high-resolution depth network merges the global 3D shape and the detailed
structures to infer the high-resolution front and back side depth maps.
Finally, an off-the-shelf mesh generator reconstructs the full 3D human model,
which are available at https://github.com/SangHunHan92/2K2K. In addition, we
also provide 2,050 3D human models, including texture maps, 3D joints, and SMPL
parameters for research purposes. In experiments, we demonstrate competitive
performance over the recent works on various datasets.
| no_new_dataset | 0.949389 |
2303.15111 | Shaozhe Hao | Shaozhe Hao, Kai Han, Kwan-Yee K. Wong | Learning Attention as Disentangler for Compositional Zero-shot Learning | CVPR 2023, available at https://haoosz.github.io/ade-czsl/ | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by/4.0/ | Compositional zero-shot learning (CZSL) aims at learning visual concepts
(i.e., attributes and objects) from seen compositions and combining concept
knowledge into unseen compositions. The key to CZSL is learning the
disentanglement of the attribute-object composition. To this end, we propose to
exploit cross-attentions as compositional disentanglers to learn disentangled
concept embeddings. For example, if we want to recognize an unseen composition
"yellow flower", we can learn the attribute concept "yellow" and object concept
"flower" from different yellow objects and different flowers respectively. To
further constrain the disentanglers to learn the concept of interest, we employ
a regularization at the attention level. Specifically, we adapt the earth
mover's distance (EMD) as a feature similarity metric in the cross-attention
module. Moreover, benefiting from concept disentanglement, we improve the
inference process and tune the prediction score by combining multiple concept
probabilities. Comprehensive experiments on three CZSL benchmark datasets
demonstrate that our method significantly outperforms previous works in both
closed- and open-world settings, establishing a new state-of-the-art.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:29:54 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Hao",
"Shaozhe",
""
],
[
"Han",
"Kai",
""
],
[
"Wong",
"Kwan-Yee K.",
""
]
] | TITLE: Learning Attention as Disentangler for Compositional Zero-shot Learning
ABSTRACT: Compositional zero-shot learning (CZSL) aims at learning visual concepts
(i.e., attributes and objects) from seen compositions and combining concept
knowledge into unseen compositions. The key to CZSL is learning the
disentanglement of the attribute-object composition. To this end, we propose to
exploit cross-attentions as compositional disentanglers to learn disentangled
concept embeddings. For example, if we want to recognize an unseen composition
"yellow flower", we can learn the attribute concept "yellow" and object concept
"flower" from different yellow objects and different flowers respectively. To
further constrain the disentanglers to learn the concept of interest, we employ
a regularization at the attention level. Specifically, we adapt the earth
mover's distance (EMD) as a feature similarity metric in the cross-attention
module. Moreover, benefiting from concept disentanglement, we improve the
inference process and tune the prediction score by combining multiple concept
probabilities. Comprehensive experiments on three CZSL benchmark datasets
demonstrate that our method significantly outperforms previous works in both
closed- and open-world settings, establishing a new state-of-the-art.
| no_new_dataset | 0.939582 |
2303.15122 | Mausoom Sarkar | Mausoom Sarkar, Nikitha SR, Mayur Hemani, Rishabh Jain, Balaji
Krishnamurthy | Parameter Efficient Local Implicit Image Function Network for Face
Segmentation | Accepted at CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Face parsing is defined as the per-pixel labeling of images containing human
faces. The labels are defined to identify key facial regions like eyes, lips,
nose, hair, etc. In this work, we make use of the structural consistency of the
human face to propose a lightweight face-parsing method using a Local Implicit
Function network, FP-LIIF. We propose a simple architecture having a
convolutional encoder and a pixel MLP decoder that uses 1/26th number of
parameters compared to the state-of-the-art models and yet matches or
outperforms state-of-the-art models on multiple datasets, like CelebAMask-HQ
and LaPa. We do not use any pretraining, and compared to other works, our
network can also generate segmentation at different resolutions without any
changes in the input resolution. This work enables the use of facial
segmentation on low-compute or low-bandwidth devices because of its higher FPS
and smaller model size.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:50:27 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Sarkar",
"Mausoom",
""
],
[
"SR",
"Nikitha",
""
],
[
"Hemani",
"Mayur",
""
],
[
"Jain",
"Rishabh",
""
],
[
"Krishnamurthy",
"Balaji",
""
]
] | TITLE: Parameter Efficient Local Implicit Image Function Network for Face
Segmentation
ABSTRACT: Face parsing is defined as the per-pixel labeling of images containing human
faces. The labels are defined to identify key facial regions like eyes, lips,
nose, hair, etc. In this work, we make use of the structural consistency of the
human face to propose a lightweight face-parsing method using a Local Implicit
Function network, FP-LIIF. We propose a simple architecture having a
convolutional encoder and a pixel MLP decoder that uses 1/26th number of
parameters compared to the state-of-the-art models and yet matches or
outperforms state-of-the-art models on multiple datasets, like CelebAMask-HQ
and LaPa. We do not use any pretraining, and compared to other works, our
network can also generate segmentation at different resolutions without any
changes in the input resolution. This work enables the use of facial
segmentation on low-compute or low-bandwidth devices because of its higher FPS
and smaller model size.
| no_new_dataset | 0.948585 |
2303.15126 | Zehan Zheng | Zehan Zheng, Danni Wu, Ruisi Lu, Fan Lu, Guang Chen, Changjun Jiang | NeuralPCI: Spatio-temporal Neural Field for 3D Point Cloud Multi-frame
Non-linear Interpolation | Accepted by CVPR 2023. Project Page:
https://dyfcalid.github.io/NeuralPCI | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In recent years, there has been a significant increase in focus on the
interpolation task of computer vision. Despite the tremendous advancement of
video interpolation, point cloud interpolation remains insufficiently explored.
Meanwhile, the existence of numerous nonlinear large motions in real-world
scenarios makes the point cloud interpolation task more challenging. In light
of these issues, we present NeuralPCI: an end-to-end 4D spatio-temporal Neural
field for 3D Point Cloud Interpolation, which implicitly integrates multi-frame
information to handle nonlinear large motions for both indoor and outdoor
scenarios. Furthermore, we construct a new multi-frame point cloud
interpolation dataset called NL-Drive for large nonlinear motions in autonomous
driving scenes to better demonstrate the superiority of our method. Ultimately,
NeuralPCI achieves state-of-the-art performance on both DHB (Dynamic Human
Bodies) and NL-Drive datasets. Beyond the interpolation task, our method can be
naturally extended to point cloud extrapolation, morphing, and auto-labeling,
which indicates its substantial potential in other domains. Codes are available
at https://github.com/ispc-lab/NeuralPCI.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 11:58:52 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Zheng",
"Zehan",
""
],
[
"Wu",
"Danni",
""
],
[
"Lu",
"Ruisi",
""
],
[
"Lu",
"Fan",
""
],
[
"Chen",
"Guang",
""
],
[
"Jiang",
"Changjun",
""
]
] | TITLE: NeuralPCI: Spatio-temporal Neural Field for 3D Point Cloud Multi-frame
Non-linear Interpolation
ABSTRACT: In recent years, there has been a significant increase in focus on the
interpolation task of computer vision. Despite the tremendous advancement of
video interpolation, point cloud interpolation remains insufficiently explored.
Meanwhile, the existence of numerous nonlinear large motions in real-world
scenarios makes the point cloud interpolation task more challenging. In light
of these issues, we present NeuralPCI: an end-to-end 4D spatio-temporal Neural
field for 3D Point Cloud Interpolation, which implicitly integrates multi-frame
information to handle nonlinear large motions for both indoor and outdoor
scenarios. Furthermore, we construct a new multi-frame point cloud
interpolation dataset called NL-Drive for large nonlinear motions in autonomous
driving scenes to better demonstrate the superiority of our method. Ultimately,
NeuralPCI achieves state-of-the-art performance on both DHB (Dynamic Human
Bodies) and NL-Drive datasets. Beyond the interpolation task, our method can be
naturally extended to point cloud extrapolation, morphing, and auto-labeling,
which indicates its substantial potential in other domains. Codes are available
at https://github.com/ispc-lab/NeuralPCI.
| new_dataset | 0.962638 |
2303.15127 | Xitong Gao | Tianrui Qin, Xitong Gao, Juanjuan Zhao, Kejiang Ye, Cheng-Zhong Xu | Learning the Unlearnable: Adversarial Augmentations Suppress Unlearnable
Example Attacks | UEraser introduces adversarial augmentations to suppress unlearnable
example attacks and outperforms current defenses | null | null | null | cs.LG cs.CR cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Unlearnable example attacks are data poisoning techniques that can be used to
safeguard public data against unauthorized use for training deep learning
models. These methods add stealthy perturbations to the original image, thereby
making it difficult for deep learning models to learn from these training data
effectively. Current research suggests that adversarial training can, to a
certain degree, mitigate the impact of unlearnable example attacks, while
common data augmentation methods are not effective against such poisons.
Adversarial training, however, demands considerable computational resources and
can result in non-trivial accuracy loss. In this paper, we introduce the
UEraser method, which outperforms current defenses against different types of
state-of-the-art unlearnable example attacks through a combination of effective
data augmentation policies and loss-maximizing adversarial augmentations. In
stark contrast to the current SOTA adversarial training methods, UEraser uses
adversarial augmentations, which extends beyond the confines of $ \ell_p $
perturbation budget assumed by current unlearning attacks and defenses. It also
helps to improve the model's generalization ability, thus protecting against
accuracy loss. UEraser wipes out the unlearning effect with error-maximizing
data augmentations, thus restoring trained model accuracies. Interestingly,
UEraser-Lite, a fast variant without adversarial augmentations, is also highly
effective in preserving clean accuracies. On challenging unlearnable CIFAR-10,
CIFAR-100, SVHN, and ImageNet-subset datasets produced with various attacks, it
achieves results that are comparable to those obtained during clean training.
We also demonstrate its efficacy against possible adaptive attacks. Our code is
open source and available to the deep learning community:
https://github.com/lafeat/ueraser.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 12:00:54 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Qin",
"Tianrui",
""
],
[
"Gao",
"Xitong",
""
],
[
"Zhao",
"Juanjuan",
""
],
[
"Ye",
"Kejiang",
""
],
[
"Xu",
"Cheng-Zhong",
""
]
] | TITLE: Learning the Unlearnable: Adversarial Augmentations Suppress Unlearnable
Example Attacks
ABSTRACT: Unlearnable example attacks are data poisoning techniques that can be used to
safeguard public data against unauthorized use for training deep learning
models. These methods add stealthy perturbations to the original image, thereby
making it difficult for deep learning models to learn from these training data
effectively. Current research suggests that adversarial training can, to a
certain degree, mitigate the impact of unlearnable example attacks, while
common data augmentation methods are not effective against such poisons.
Adversarial training, however, demands considerable computational resources and
can result in non-trivial accuracy loss. In this paper, we introduce the
UEraser method, which outperforms current defenses against different types of
state-of-the-art unlearnable example attacks through a combination of effective
data augmentation policies and loss-maximizing adversarial augmentations. In
stark contrast to the current SOTA adversarial training methods, UEraser uses
adversarial augmentations, which extends beyond the confines of $ \ell_p $
perturbation budget assumed by current unlearning attacks and defenses. It also
helps to improve the model's generalization ability, thus protecting against
accuracy loss. UEraser wipes out the unlearning effect with error-maximizing
data augmentations, thus restoring trained model accuracies. Interestingly,
UEraser-Lite, a fast variant without adversarial augmentations, is also highly
effective in preserving clean accuracies. On challenging unlearnable CIFAR-10,
CIFAR-100, SVHN, and ImageNet-subset datasets produced with various attacks, it
achieves results that are comparable to those obtained during clean training.
We also demonstrate its efficacy against possible adaptive attacks. Our code is
open source and available to the deep learning community:
https://github.com/lafeat/ueraser.
| no_new_dataset | 0.945248 |
2303.15143 | He Fang | He Fang, Zhenlong Xiao, Xianbin Wang, Li Xu, and Lajos Hanzo | Collaborative Authentication for 6G Networks: An Edge Intelligence based
Autonomous Approach | null | null | null | null | cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The conventional device authentication of wireless networks usually relies on
a security server and centralized process, leading to long latency and risk of
single-point of failure. While these challenges might be mitigated by
collaborative authentication schemes, their performance remains limited by the
rigidity of data collection and aggregated result. They also tend to ignore
attacker localization in the collaborative authentication process. To overcome
these challenges, a novel collaborative authentication scheme is proposed,
where multiple edge devices act as cooperative peers to assist the service
provider in distributively authenticating its users by estimating their
received signal strength indicator (RSSI) and mobility trajectory (TRA). More
explicitly, a distributed learning-based collaborative authentication algorithm
is conceived, where the cooperative peers update their authentication models
locally, thus the network congestion and response time remain low. Moreover, a
situation-aware secure group update algorithm is proposed for autonomously
refreshing the set of cooperative peers in the dynamic environment. We also
develop an algorithm for localizing a malicious user by the cooperative peers
once it is identified. The simulation results demonstrate that the proposed
scheme is eminently suitable for both indoor and outdoor communication
scenarios, and outperforms some existing benchmark schemes.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 12:22:41 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Fang",
"He",
""
],
[
"Xiao",
"Zhenlong",
""
],
[
"Wang",
"Xianbin",
""
],
[
"Xu",
"Li",
""
],
[
"Hanzo",
"Lajos",
""
]
] | TITLE: Collaborative Authentication for 6G Networks: An Edge Intelligence based
Autonomous Approach
ABSTRACT: The conventional device authentication of wireless networks usually relies on
a security server and centralized process, leading to long latency and risk of
single-point of failure. While these challenges might be mitigated by
collaborative authentication schemes, their performance remains limited by the
rigidity of data collection and aggregated result. They also tend to ignore
attacker localization in the collaborative authentication process. To overcome
these challenges, a novel collaborative authentication scheme is proposed,
where multiple edge devices act as cooperative peers to assist the service
provider in distributively authenticating its users by estimating their
received signal strength indicator (RSSI) and mobility trajectory (TRA). More
explicitly, a distributed learning-based collaborative authentication algorithm
is conceived, where the cooperative peers update their authentication models
locally, thus the network congestion and response time remain low. Moreover, a
situation-aware secure group update algorithm is proposed for autonomously
refreshing the set of cooperative peers in the dynamic environment. We also
develop an algorithm for localizing a malicious user by the cooperative peers
once it is identified. The simulation results demonstrate that the proposed
scheme is eminently suitable for both indoor and outdoor communication
scenarios, and outperforms some existing benchmark schemes.
| no_new_dataset | 0.94428 |
2303.15166 | Haoyuan Tian | Ran Yi, Haoyuan Tian, Zhihao Gu, Yu-Kun Lai and Paul L. Rosin | Towards Artistic Image Aesthetics Assessment: a Large-scale Dataset and
a New Method | Accepted by CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Image aesthetics assessment (IAA) is a challenging task due to its highly
subjective nature. Most of the current studies rely on large-scale datasets
(e.g., AVA and AADB) to learn a general model for all kinds of photography
images. However, little light has been shed on measuring the aesthetic quality
of artistic images, and the existing datasets only contain relatively few
artworks. Such a defect is a great obstacle to the aesthetic assessment of
artistic images. To fill the gap in the field of artistic image aesthetics
assessment (AIAA), we first introduce a large-scale AIAA dataset: Boldbrush
Artistic Image Dataset (BAID), which consists of 60,337 artistic images
covering various art forms, with more than 360,000 votes from online users. We
then propose a new method, SAAN (Style-specific Art Assessment Network), which
can effectively extract and utilize style-specific and generic aesthetic
information to evaluate artistic images. Experiments demonstrate that our
proposed approach outperforms existing IAA methods on the proposed BAID dataset
according to quantitative comparisons. We believe the proposed dataset and
method can serve as a foundation for future AIAA works and inspire more
research in this field. Dataset and code are available at:
https://github.com/Dreemurr-T/BAID.git
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 12:59:15 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Yi",
"Ran",
""
],
[
"Tian",
"Haoyuan",
""
],
[
"Gu",
"Zhihao",
""
],
[
"Lai",
"Yu-Kun",
""
],
[
"Rosin",
"Paul L.",
""
]
] | TITLE: Towards Artistic Image Aesthetics Assessment: a Large-scale Dataset and
a New Method
ABSTRACT: Image aesthetics assessment (IAA) is a challenging task due to its highly
subjective nature. Most of the current studies rely on large-scale datasets
(e.g., AVA and AADB) to learn a general model for all kinds of photography
images. However, little light has been shed on measuring the aesthetic quality
of artistic images, and the existing datasets only contain relatively few
artworks. Such a defect is a great obstacle to the aesthetic assessment of
artistic images. To fill the gap in the field of artistic image aesthetics
assessment (AIAA), we first introduce a large-scale AIAA dataset: Boldbrush
Artistic Image Dataset (BAID), which consists of 60,337 artistic images
covering various art forms, with more than 360,000 votes from online users. We
then propose a new method, SAAN (Style-specific Art Assessment Network), which
can effectively extract and utilize style-specific and generic aesthetic
information to evaluate artistic images. Experiments demonstrate that our
proposed approach outperforms existing IAA methods on the proposed BAID dataset
according to quantitative comparisons. We believe the proposed dataset and
method can serve as a foundation for future AIAA works and inspire more
research in this field. Dataset and code are available at:
https://github.com/Dreemurr-T/BAID.git
| new_dataset | 0.964119 |
2303.15167 | Fumiaki Sato | Fumiaki Sato, Ryo Hachiuma, Taiki Sekii | Prompt-Guided Zero-Shot Anomaly Action Recognition using Pretrained Deep
Skeleton Features | CVPR 2023 | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | This study investigates unsupervised anomaly action recognition, which
identifies video-level abnormal-human-behavior events in an unsupervised manner
without abnormal samples, and simultaneously addresses three limitations in the
conventional skeleton-based approaches: target domain-dependent DNN training,
robustness against skeleton errors, and a lack of normal samples. We present a
unified, user prompt-guided zero-shot learning framework using a target
domain-independent skeleton feature extractor, which is pretrained on a
large-scale action recognition dataset. Particularly, during the training phase
using normal samples, the method models the distribution of skeleton features
of the normal actions while freezing the weights of the DNNs and estimates the
anomaly score using this distribution in the inference phase. Additionally, to
increase robustness against skeleton errors, we introduce a DNN architecture
inspired by a point cloud deep learning paradigm, which sparsely propagates the
features between joints. Furthermore, to prevent the unobserved normal actions
from being misidentified as abnormal actions, we incorporate a similarity score
between the user prompt embeddings and skeleton features aligned in the common
space into the anomaly score, which indirectly supplements normal actions. On
two publicly available datasets, we conduct experiments to test the
effectiveness of the proposed method with respect to abovementioned
limitations.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 12:59:33 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Sato",
"Fumiaki",
""
],
[
"Hachiuma",
"Ryo",
""
],
[
"Sekii",
"Taiki",
""
]
] | TITLE: Prompt-Guided Zero-Shot Anomaly Action Recognition using Pretrained Deep
Skeleton Features
ABSTRACT: This study investigates unsupervised anomaly action recognition, which
identifies video-level abnormal-human-behavior events in an unsupervised manner
without abnormal samples, and simultaneously addresses three limitations in the
conventional skeleton-based approaches: target domain-dependent DNN training,
robustness against skeleton errors, and a lack of normal samples. We present a
unified, user prompt-guided zero-shot learning framework using a target
domain-independent skeleton feature extractor, which is pretrained on a
large-scale action recognition dataset. Particularly, during the training phase
using normal samples, the method models the distribution of skeleton features
of the normal actions while freezing the weights of the DNNs and estimates the
anomaly score using this distribution in the inference phase. Additionally, to
increase robustness against skeleton errors, we introduce a DNN architecture
inspired by a point cloud deep learning paradigm, which sparsely propagates the
features between joints. Furthermore, to prevent the unobserved normal actions
from being misidentified as abnormal actions, we incorporate a similarity score
between the user prompt embeddings and skeleton features aligned in the common
space into the anomaly score, which indirectly supplements normal actions. On
two publicly available datasets, we conduct experiments to test the
effectiveness of the proposed method with respect to abovementioned
limitations.
| no_new_dataset | 0.94743 |
2303.15168 | Yang Lu | Yang Lu, Pinxin Qian, Gang Huang, Hanzi Wang | Personalized Federated Learning on Long-Tailed Data via Adversarial
Feature Augmentation | Accepted by ICASSP 2023 | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Personalized Federated Learning (PFL) aims to learn personalized models for
each client based on the knowledge across all clients in a privacy-preserving
manner. Existing PFL methods generally assume that the underlying global data
across all clients are uniformly distributed without considering the long-tail
distribution. The joint problem of data heterogeneity and long-tail
distribution in the FL environment is more challenging and severely affects the
performance of personalized models. In this paper, we propose a PFL method
called Federated Learning with Adversarial Feature Augmentation (FedAFA) to
address this joint problem in PFL. FedAFA optimizes the personalized model for
each client by producing a balanced feature set to enhance the local minority
classes. The local minority class features are generated by transferring the
knowledge from the local majority class features extracted by the global model
in an adversarial example learning manner. The experimental results on
benchmarks under different settings of data heterogeneity and long-tail
distribution demonstrate that FedAFA significantly improves the personalized
performance of each client compared with the state-of-the-art PFL algorithm.
The code is available at https://github.com/pxqian/FedAFA.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 13:00:20 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Lu",
"Yang",
""
],
[
"Qian",
"Pinxin",
""
],
[
"Huang",
"Gang",
""
],
[
"Wang",
"Hanzi",
""
]
] | TITLE: Personalized Federated Learning on Long-Tailed Data via Adversarial
Feature Augmentation
ABSTRACT: Personalized Federated Learning (PFL) aims to learn personalized models for
each client based on the knowledge across all clients in a privacy-preserving
manner. Existing PFL methods generally assume that the underlying global data
across all clients are uniformly distributed without considering the long-tail
distribution. The joint problem of data heterogeneity and long-tail
distribution in the FL environment is more challenging and severely affects the
performance of personalized models. In this paper, we propose a PFL method
called Federated Learning with Adversarial Feature Augmentation (FedAFA) to
address this joint problem in PFL. FedAFA optimizes the personalized model for
each client by producing a balanced feature set to enhance the local minority
classes. The local minority class features are generated by transferring the
knowledge from the local majority class features extracted by the global model
in an adversarial example learning manner. The experimental results on
benchmarks under different settings of data heterogeneity and long-tail
distribution demonstrate that FedAFA significantly improves the personalized
performance of each client compared with the state-of-the-art PFL algorithm.
The code is available at https://github.com/pxqian/FedAFA.
| no_new_dataset | 0.949201 |
2303.15180 | Shiwei Feng | Shiwei Feng, Guanhong Tao, Siyuan Cheng, Guangyu Shen, Xiangzhe Xu,
Yingqi Liu, Kaiyuan Zhang, Shiqing Ma, Xiangyu Zhang | Detecting Backdoors in Pre-trained Encoders | Accepted at CVPR 2023. Code is available at
https://github.com/GiantSeaweed/DECREE | null | null | null | cs.CV cs.AI cs.CR | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Self-supervised learning in computer vision trains on unlabeled data, such as
images or (image, text) pairs, to obtain an image encoder that learns
high-quality embeddings for input data. Emerging backdoor attacks towards
encoders expose crucial vulnerabilities of self-supervised learning, since
downstream classifiers (even further trained on clean data) may inherit
backdoor behaviors from encoders. Existing backdoor detection methods mainly
focus on supervised learning settings and cannot handle pre-trained encoders
especially when input labels are not available. In this paper, we propose
DECREE, the first backdoor detection approach for pre-trained encoders,
requiring neither classifier headers nor input labels. We evaluate DECREE on
over 400 encoders trojaned under 3 paradigms. We show the effectiveness of our
method on image encoders pre-trained on ImageNet and OpenAI's CLIP 400 million
image-text pairs. Our method consistently has a high detection accuracy even if
we have only limited or no access to the pre-training dataset.
| [
{
"version": "v1",
"created": "Thu, 23 Mar 2023 19:04:40 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Feng",
"Shiwei",
""
],
[
"Tao",
"Guanhong",
""
],
[
"Cheng",
"Siyuan",
""
],
[
"Shen",
"Guangyu",
""
],
[
"Xu",
"Xiangzhe",
""
],
[
"Liu",
"Yingqi",
""
],
[
"Zhang",
"Kaiyuan",
""
],
[
"Ma",
"Shiqing",
""
],
[
"Zhang",
"Xiangyu",
""
]
] | TITLE: Detecting Backdoors in Pre-trained Encoders
ABSTRACT: Self-supervised learning in computer vision trains on unlabeled data, such as
images or (image, text) pairs, to obtain an image encoder that learns
high-quality embeddings for input data. Emerging backdoor attacks towards
encoders expose crucial vulnerabilities of self-supervised learning, since
downstream classifiers (even further trained on clean data) may inherit
backdoor behaviors from encoders. Existing backdoor detection methods mainly
focus on supervised learning settings and cannot handle pre-trained encoders
especially when input labels are not available. In this paper, we propose
DECREE, the first backdoor detection approach for pre-trained encoders,
requiring neither classifier headers nor input labels. We evaluate DECREE on
over 400 encoders trojaned under 3 paradigms. We show the effectiveness of our
method on image encoders pre-trained on ImageNet and OpenAI's CLIP 400 million
image-text pairs. Our method consistently has a high detection accuracy even if
we have only limited or no access to the pre-training dataset.
| no_new_dataset | 0.943191 |
2303.15218 | Gissel Velarde | Gissel Velarde, Anindya Sudhir, Sanjay Deshmane, Anuj Deshmunkh,
Khushboo Sharma and Vaibhav Joshi | Evaluating XGBoost for Balanced and Imbalanced Data: Application to
Fraud Detection | 17 pages, 8 figures, 9 tables, Presented at NVIDIA GTC, The
Conference for the Era of AI and the Metaverse, March 23, 2023. [S51129] | null | null | null | cs.LG cs.AI | http://creativecommons.org/licenses/by/4.0/ | This paper evaluates XGboost's performance given different dataset sizes and
class distributions, from perfectly balanced to highly imbalanced. XGBoost has
been selected for evaluation, as it stands out in several benchmarks due to its
detection performance and speed. After introducing the problem of fraud
detection, the paper reviews evaluation metrics for detection systems or binary
classifiers, and illustrates with examples how different metrics work for
balanced and imbalanced datasets. Then, it examines the principles of XGBoost.
It proposes a pipeline for data preparation and compares a Vanilla XGBoost
against a random search-tuned XGBoost. Random search fine-tuning provides
consistent improvement for large datasets of 100 thousand samples, not so for
medium and small datasets of 10 and 1 thousand samples, respectively. Besides,
as expected, XGBoost recognition performance improves as more data is
available, and deteriorates detection performance as the datasets become more
imbalanced. Tests on distributions with 50, 45, 25, and 5 percent positive
samples show that the largest drop in detection performance occurs for the
distribution with only 5 percent positive samples. Sampling to balance the
training set does not provide consistent improvement. Therefore, future work
will include a systematic study of different techniques to deal with data
imbalance and evaluating other approaches, including graphs, autoencoders, and
generative adversarial methods, to deal with the lack of labels.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 13:59:22 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Velarde",
"Gissel",
""
],
[
"Sudhir",
"Anindya",
""
],
[
"Deshmane",
"Sanjay",
""
],
[
"Deshmunkh",
"Anuj",
""
],
[
"Sharma",
"Khushboo",
""
],
[
"Joshi",
"Vaibhav",
""
]
] | TITLE: Evaluating XGBoost for Balanced and Imbalanced Data: Application to
Fraud Detection
ABSTRACT: This paper evaluates XGboost's performance given different dataset sizes and
class distributions, from perfectly balanced to highly imbalanced. XGBoost has
been selected for evaluation, as it stands out in several benchmarks due to its
detection performance and speed. After introducing the problem of fraud
detection, the paper reviews evaluation metrics for detection systems or binary
classifiers, and illustrates with examples how different metrics work for
balanced and imbalanced datasets. Then, it examines the principles of XGBoost.
It proposes a pipeline for data preparation and compares a Vanilla XGBoost
against a random search-tuned XGBoost. Random search fine-tuning provides
consistent improvement for large datasets of 100 thousand samples, not so for
medium and small datasets of 10 and 1 thousand samples, respectively. Besides,
as expected, XGBoost recognition performance improves as more data is
available, and deteriorates detection performance as the datasets become more
imbalanced. Tests on distributions with 50, 45, 25, and 5 percent positive
samples show that the largest drop in detection performance occurs for the
distribution with only 5 percent positive samples. Sampling to balance the
training set does not provide consistent improvement. Therefore, future work
will include a systematic study of different techniques to deal with data
imbalance and evaluating other approaches, including graphs, autoencoders, and
generative adversarial methods, to deal with the lack of labels.
| no_new_dataset | 0.941815 |
2303.15219 | Eli Friedman | Eli Friedman, Assaf Lehr, Alexey Gruzdev, Vladimir Loginov, Max Kogan,
Moran Rubin, Orly Zvitia | Knowing the Distance: Understanding the Gap Between Synthetic and Real
Data For Face Parsing | null | null | null | null | cs.CV cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | The use of synthetic data for training computer vision algorithms has become
increasingly popular due to its cost-effectiveness, scalability, and ability to
provide accurate multi-modality labels. Although recent studies have
demonstrated impressive results when training networks solely on synthetic
data, there remains a performance gap between synthetic and real data that is
commonly attributed to lack of photorealism. The aim of this study is to
investigate the gap in greater detail for the face parsing task. We
differentiate between three types of gaps: distribution gap, label gap, and
photorealism gap. Our findings show that the distribution gap is the largest
contributor to the performance gap, accounting for over 50% of the gap. By
addressing this gap and accounting for the labels gap, we demonstrate that a
model trained on synthetic data achieves comparable results to one trained on a
similar amount of real data. This suggests that synthetic data is a viable
alternative to real data, especially when real data is limited or difficult to
obtain. Our study highlights the importance of content diversity in synthetic
datasets and challenges the notion that the photorealism gap is the most
critical factor affecting the performance of computer vision models trained on
synthetic data.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 13:59:26 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Friedman",
"Eli",
""
],
[
"Lehr",
"Assaf",
""
],
[
"Gruzdev",
"Alexey",
""
],
[
"Loginov",
"Vladimir",
""
],
[
"Kogan",
"Max",
""
],
[
"Rubin",
"Moran",
""
],
[
"Zvitia",
"Orly",
""
]
] | TITLE: Knowing the Distance: Understanding the Gap Between Synthetic and Real
Data For Face Parsing
ABSTRACT: The use of synthetic data for training computer vision algorithms has become
increasingly popular due to its cost-effectiveness, scalability, and ability to
provide accurate multi-modality labels. Although recent studies have
demonstrated impressive results when training networks solely on synthetic
data, there remains a performance gap between synthetic and real data that is
commonly attributed to lack of photorealism. The aim of this study is to
investigate the gap in greater detail for the face parsing task. We
differentiate between three types of gaps: distribution gap, label gap, and
photorealism gap. Our findings show that the distribution gap is the largest
contributor to the performance gap, accounting for over 50% of the gap. By
addressing this gap and accounting for the labels gap, we demonstrate that a
model trained on synthetic data achieves comparable results to one trained on a
similar amount of real data. This suggests that synthetic data is a viable
alternative to real data, especially when real data is limited or difficult to
obtain. Our study highlights the importance of content diversity in synthetic
datasets and challenges the notion that the photorealism gap is the most
critical factor affecting the performance of computer vision models trained on
synthetic data.
| no_new_dataset | 0.950641 |
2303.15223 | Sayeh Gholipour Picha | Sayeh Gholipour Picha, Dawood AL Chanti, Alice Caplier | How far generated data can impact Neural Networks performance? | Conference Publication in Proceedings of the 18th International Joint
Conference on Computer Vision, Imaging and Computer Graphics Theory and
Applications - Volume 5: VISAPP, 10 pages | null | 10.5220/0011629000003417 | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | The success of deep learning models depends on the size and quality of the
dataset to solve certain tasks. Here, we explore how far generated data can aid
real data in improving the performance of Neural Networks. In this work, we
consider facial expression recognition since it requires challenging local data
generation at the level of local regions such as mouth, eyebrows, etc, rather
than simple augmentation. Generative Adversarial Networks (GANs) provide an
alternative method for generating such local deformations but they need further
validation. To answer our question, we consider noncomplex Convolutional Neural
Networks (CNNs) based classifiers for recognizing Ekman emotions. For the data
generation process, we consider generating facial expressions (FEs) by relying
on two GANs. The first generates a random identity while the second imposes
facial deformations on top of it. We consider training the CNN classifier using
FEs from: real-faces, GANs-generated, and finally using a combination of real
and GAN-generated faces. We determine an upper bound regarding the data
generation quantity to be mixed with the real one which contributes the most to
enhancing FER accuracy. In our experiments, we find out that 5-times more
synthetic data to the real FEs dataset increases accuracy by 16%.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 14:02:43 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Picha",
"Sayeh Gholipour",
""
],
[
"Chanti",
"Dawood AL",
""
],
[
"Caplier",
"Alice",
""
]
] | TITLE: How far generated data can impact Neural Networks performance?
ABSTRACT: The success of deep learning models depends on the size and quality of the
dataset to solve certain tasks. Here, we explore how far generated data can aid
real data in improving the performance of Neural Networks. In this work, we
consider facial expression recognition since it requires challenging local data
generation at the level of local regions such as mouth, eyebrows, etc, rather
than simple augmentation. Generative Adversarial Networks (GANs) provide an
alternative method for generating such local deformations but they need further
validation. To answer our question, we consider noncomplex Convolutional Neural
Networks (CNNs) based classifiers for recognizing Ekman emotions. For the data
generation process, we consider generating facial expressions (FEs) by relying
on two GANs. The first generates a random identity while the second imposes
facial deformations on top of it. We consider training the CNN classifier using
FEs from: real-faces, GANs-generated, and finally using a combination of real
and GAN-generated faces. We determine an upper bound regarding the data
generation quantity to be mixed with the real one which contributes the most to
enhancing FER accuracy. In our experiments, we find out that 5-times more
synthetic data to the real FEs dataset increases accuracy by 16%.
| no_new_dataset | 0.951006 |
2303.15224 | Guangzhi Tang | Guangzhi Tang, Ali Safa, Kevin Shidqi, Paul Detterer, Stefano
Traferro, Mario Konijnenburg, Manolis Sifalakis, Gert-Jan van Schaik,
Amirreza Yousefzadeh | Open the box of digital neuromorphic processor: Towards effective
algorithm-hardware co-design | null | null | null | null | cs.NE cs.SY eess.SY | http://creativecommons.org/licenses/by/4.0/ | Sparse and event-driven spiking neural network (SNN) algorithms are the ideal
candidate solution for energy-efficient edge computing. Yet, with the growing
complexity of SNN algorithms, it isn't easy to properly benchmark and optimize
their computational cost without hardware in the loop. Although digital
neuromorphic processors have been widely adopted to benchmark SNN algorithms,
their black-box nature is problematic for algorithm-hardware co-optimization.
In this work, we open the black box of the digital neuromorphic processor for
algorithm designers by presenting the neuron processing instruction set and
detailed energy consumption of the SENeCA neuromorphic architecture. For
convenient benchmarking and optimization, we provide the energy cost of the
essential neuromorphic components in SENeCA, including neuron models and
learning rules. Moreover, we exploit the SENeCA's hierarchical memory and
exhibit an advantage over existing neuromorphic processors. We show the energy
efficiency of SNN algorithms for video processing and online learning, and
demonstrate the potential of our work for optimizing algorithm designs.
Overall, we present a practical approach to enable algorithm designers to
accurately benchmark SNN algorithms and pave the way towards effective
algorithm-hardware co-design.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 14:03:11 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Tang",
"Guangzhi",
""
],
[
"Safa",
"Ali",
""
],
[
"Shidqi",
"Kevin",
""
],
[
"Detterer",
"Paul",
""
],
[
"Traferro",
"Stefano",
""
],
[
"Konijnenburg",
"Mario",
""
],
[
"Sifalakis",
"Manolis",
""
],
[
"van Schaik",
"Gert-Jan",
""
],
[
"Yousefzadeh",
"Amirreza",
""
]
] | TITLE: Open the box of digital neuromorphic processor: Towards effective
algorithm-hardware co-design
ABSTRACT: Sparse and event-driven spiking neural network (SNN) algorithms are the ideal
candidate solution for energy-efficient edge computing. Yet, with the growing
complexity of SNN algorithms, it isn't easy to properly benchmark and optimize
their computational cost without hardware in the loop. Although digital
neuromorphic processors have been widely adopted to benchmark SNN algorithms,
their black-box nature is problematic for algorithm-hardware co-optimization.
In this work, we open the black box of the digital neuromorphic processor for
algorithm designers by presenting the neuron processing instruction set and
detailed energy consumption of the SENeCA neuromorphic architecture. For
convenient benchmarking and optimization, we provide the energy cost of the
essential neuromorphic components in SENeCA, including neuron models and
learning rules. Moreover, we exploit the SENeCA's hierarchical memory and
exhibit an advantage over existing neuromorphic processors. We show the energy
efficiency of SNN algorithms for video processing and online learning, and
demonstrate the potential of our work for optimizing algorithm designs.
Overall, we present a practical approach to enable algorithm designers to
accurately benchmark SNN algorithms and pave the way towards effective
algorithm-hardware co-design.
| no_new_dataset | 0.94743 |
2303.15234 | Jingchen Sun | Jingchen Sun, Jiayu Qin, Zihao Lin, Changyou Chen | Prompt Tuning based Adapter for Vision-Language Model Adaption | null | null | null | null | cs.CV cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large pre-trained vision-language (VL) models have shown significant promise
in adapting to various downstream tasks. However, fine-tuning the entire
network is challenging due to the massive number of model parameters. To
address this issue, efficient adaptation methods such as prompt tuning have
been proposed. We explore the idea of prompt tuning with multi-task pre-trained
initialization and find it can significantly improve model performance. Based
on our findings, we introduce a new model, termed Prompt-Adapter, that combines
pre-trained prompt tunning with an efficient adaptation network. Our approach
beat the state-of-the-art methods in few-shot image classification on the
public 11 datasets, especially in settings with limited data instances such as
1 shot, 2 shots, 4 shots, and 8 shots images. Our proposed method demonstrates
the promise of combining prompt tuning and parameter-efficient networks for
efficient vision-language model adaptation. The code is publicly available at:
https://github.com/Jingchensun/prompt_adapter.
| [
{
"version": "v1",
"created": "Fri, 24 Mar 2023 15:05:17 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Sun",
"Jingchen",
""
],
[
"Qin",
"Jiayu",
""
],
[
"Lin",
"Zihao",
""
],
[
"Chen",
"Changyou",
""
]
] | TITLE: Prompt Tuning based Adapter for Vision-Language Model Adaption
ABSTRACT: Large pre-trained vision-language (VL) models have shown significant promise
in adapting to various downstream tasks. However, fine-tuning the entire
network is challenging due to the massive number of model parameters. To
address this issue, efficient adaptation methods such as prompt tuning have
been proposed. We explore the idea of prompt tuning with multi-task pre-trained
initialization and find it can significantly improve model performance. Based
on our findings, we introduce a new model, termed Prompt-Adapter, that combines
pre-trained prompt tunning with an efficient adaptation network. Our approach
beat the state-of-the-art methods in few-shot image classification on the
public 11 datasets, especially in settings with limited data instances such as
1 shot, 2 shots, 4 shots, and 8 shots images. Our proposed method demonstrates
the promise of combining prompt tuning and parameter-efficient networks for
efficient vision-language model adaptation. The code is publicly available at:
https://github.com/Jingchensun/prompt_adapter.
| no_new_dataset | 0.949949 |
2303.15254 | Lisa Gaedke-Merzh\"auser | Lisa Gaedke-Merzh\"auser, Elias Krainski, Radim Janalik, H{\aa}vard
Rue, Olaf Schenk | Integrated Nested Laplace Approximations for Large-Scale
Spatial-Temporal Bayesian Modeling | 22 pages, 14 figures | null | null | null | stat.CO cs.DC cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Bayesian inference tasks continue to pose a computational challenge. This
especially holds for spatial-temporal modeling where high-dimensional latent
parameter spaces are ubiquitous. The methodology of integrated nested Laplace
approximations (INLA) provides a framework for performing Bayesian inference
applicable to a large subclass of additive Bayesian hierarchical models. In
combination with the stochastic partial differential equations (SPDE) approach
it gives rise to an efficient method for spatial-temporal modeling. In this
work we build on the INLA-SPDE approach, by putting forward a performant
distributed memory variant, INLA-DIST, for large-scale applications. To perform
the arising computational kernel operations, consisting of Cholesky
factorizations, solving linear systems, and selected matrix inversions, we
present two numerical solver options, a sparse CPU-based library and a novel
blocked GPU-accelerated approach which we propose. We leverage the recurring
nonzero block structure in the arising precision (inverse covariance) matrices,
which allows us to employ dense subroutines within a sparse setting. Both
versions of INLA-DIST are highly scalable, capable of performing inference on
models with millions of latent parameters. We demonstrate their accuracy and
performance on synthetic as well as real-world climate dataset applications.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 14:42:49 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Gaedke-Merzhäuser",
"Lisa",
""
],
[
"Krainski",
"Elias",
""
],
[
"Janalik",
"Radim",
""
],
[
"Rue",
"Håvard",
""
],
[
"Schenk",
"Olaf",
""
]
] | TITLE: Integrated Nested Laplace Approximations for Large-Scale
Spatial-Temporal Bayesian Modeling
ABSTRACT: Bayesian inference tasks continue to pose a computational challenge. This
especially holds for spatial-temporal modeling where high-dimensional latent
parameter spaces are ubiquitous. The methodology of integrated nested Laplace
approximations (INLA) provides a framework for performing Bayesian inference
applicable to a large subclass of additive Bayesian hierarchical models. In
combination with the stochastic partial differential equations (SPDE) approach
it gives rise to an efficient method for spatial-temporal modeling. In this
work we build on the INLA-SPDE approach, by putting forward a performant
distributed memory variant, INLA-DIST, for large-scale applications. To perform
the arising computational kernel operations, consisting of Cholesky
factorizations, solving linear systems, and selected matrix inversions, we
present two numerical solver options, a sparse CPU-based library and a novel
blocked GPU-accelerated approach which we propose. We leverage the recurring
nonzero block structure in the arising precision (inverse covariance) matrices,
which allows us to employ dense subroutines within a sparse setting. Both
versions of INLA-DIST are highly scalable, capable of performing inference on
models with millions of latent parameters. We demonstrate their accuracy and
performance on synthetic as well as real-world climate dataset applications.
| no_new_dataset | 0.941385 |
2303.15269 | Silvia Cascianelli PhD | Vittorio Pippi, Silvia Cascianelli, Rita Cucchiara | Handwritten Text Generation from Visual Archetypes | Accepted at CVPR2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Generating synthetic images of handwritten text in a writer-specific style is
a challenging task, especially in the case of unseen styles and new words, and
even more when these latter contain characters that are rarely encountered
during training. While emulating a writer's style has been recently addressed
by generative models, the generalization towards rare characters has been
disregarded. In this work, we devise a Transformer-based model for Few-Shot
styled handwritten text generation and focus on obtaining a robust and
informative representation of both the text and the style. In particular, we
propose a novel representation of the textual content as a sequence of dense
vectors obtained from images of symbols written as standard GNU Unifont glyphs,
which can be considered their visual archetypes. This strategy is more suitable
for generating characters that, despite having been seen rarely during
training, possibly share visual details with the frequently observed ones. As
for the style, we obtain a robust representation of unseen writers' calligraphy
by exploiting specific pre-training on a large synthetic dataset. Quantitative
and qualitative results demonstrate the effectiveness of our proposal in
generating words in unseen styles and with rare characters more faithfully than
existing approaches relying on independent one-hot encodings of the characters.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 14:58:20 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Pippi",
"Vittorio",
""
],
[
"Cascianelli",
"Silvia",
""
],
[
"Cucchiara",
"Rita",
""
]
] | TITLE: Handwritten Text Generation from Visual Archetypes
ABSTRACT: Generating synthetic images of handwritten text in a writer-specific style is
a challenging task, especially in the case of unseen styles and new words, and
even more when these latter contain characters that are rarely encountered
during training. While emulating a writer's style has been recently addressed
by generative models, the generalization towards rare characters has been
disregarded. In this work, we devise a Transformer-based model for Few-Shot
styled handwritten text generation and focus on obtaining a robust and
informative representation of both the text and the style. In particular, we
propose a novel representation of the textual content as a sequence of dense
vectors obtained from images of symbols written as standard GNU Unifont glyphs,
which can be considered their visual archetypes. This strategy is more suitable
for generating characters that, despite having been seen rarely during
training, possibly share visual details with the frequently observed ones. As
for the style, we obtain a robust representation of unseen writers' calligraphy
by exploiting specific pre-training on a large synthetic dataset. Quantitative
and qualitative results demonstrate the effectiveness of our proposal in
generating words in unseen styles and with rare characters more faithfully than
existing approaches relying on independent one-hot encodings of the characters.
| no_new_dataset | 0.940353 |
2303.15286 | Yurong You | Yurong You, Cheng Perng Phoo, Katie Z Luo, Travis Zhang, Wei-Lun Chao,
Bharath Hariharan, Mark Campbell, Kilian Q. Weinberger | Unsupervised Adaptation from Repeated Traversals for Autonomous Driving | Accepted by NeurIPS 2022. Code is available at
https://github.com/YurongYou/Rote-DA | null | null | null | cs.CV cs.LG | http://creativecommons.org/licenses/by/4.0/ | For a self-driving car to operate reliably, its perceptual system must
generalize to the end-user's environment -- ideally without additional
annotation efforts. One potential solution is to leverage unlabeled data (e.g.,
unlabeled LiDAR point clouds) collected from the end-users' environments (i.e.
target domain) to adapt the system to the difference between training and
testing environments. While extensive research has been done on such an
unsupervised domain adaptation problem, one fundamental problem lingers: there
is no reliable signal in the target domain to supervise the adaptation process.
To overcome this issue we observe that it is easy to collect unsupervised data
from multiple traversals of repeated routes. While different from conventional
unsupervised domain adaptation, this assumption is extremely realistic since
many drivers share the same roads. We show that this simple additional
assumption is sufficient to obtain a potent signal that allows us to perform
iterative self-training of 3D object detectors on the target domain.
Concretely, we generate pseudo-labels with the out-of-domain detector but
reduce false positives by removing detections of supposedly mobile objects that
are persistent across traversals. Further, we reduce false negatives by
encouraging predictions in regions that are not persistent. We experiment with
our approach on two large-scale driving datasets and show remarkable
improvement in 3D object detection of cars, pedestrians, and cyclists, bringing
us a step closer to generalizable autonomous driving.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 15:07:55 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"You",
"Yurong",
""
],
[
"Phoo",
"Cheng Perng",
""
],
[
"Luo",
"Katie Z",
""
],
[
"Zhang",
"Travis",
""
],
[
"Chao",
"Wei-Lun",
""
],
[
"Hariharan",
"Bharath",
""
],
[
"Campbell",
"Mark",
""
],
[
"Weinberger",
"Kilian Q.",
""
]
] | TITLE: Unsupervised Adaptation from Repeated Traversals for Autonomous Driving
ABSTRACT: For a self-driving car to operate reliably, its perceptual system must
generalize to the end-user's environment -- ideally without additional
annotation efforts. One potential solution is to leverage unlabeled data (e.g.,
unlabeled LiDAR point clouds) collected from the end-users' environments (i.e.
target domain) to adapt the system to the difference between training and
testing environments. While extensive research has been done on such an
unsupervised domain adaptation problem, one fundamental problem lingers: there
is no reliable signal in the target domain to supervise the adaptation process.
To overcome this issue we observe that it is easy to collect unsupervised data
from multiple traversals of repeated routes. While different from conventional
unsupervised domain adaptation, this assumption is extremely realistic since
many drivers share the same roads. We show that this simple additional
assumption is sufficient to obtain a potent signal that allows us to perform
iterative self-training of 3D object detectors on the target domain.
Concretely, we generate pseudo-labels with the out-of-domain detector but
reduce false positives by removing detections of supposedly mobile objects that
are persistent across traversals. Further, we reduce false negatives by
encouraging predictions in regions that are not persistent. We experiment with
our approach on two large-scale driving datasets and show remarkable
improvement in 3D object detection of cars, pedestrians, and cyclists, bringing
us a step closer to generalizable autonomous driving.
| no_new_dataset | 0.940079 |
2303.15291 | Bhargav Joshi | Bhargav Joshi and Taihui Li and Buyun Liang and Roger Rusack and Ju
Sun | Predicting the Future of the CMS Detector: Crystal Radiation Damage and
Machine Learning at the LHC | null | null | null | null | physics.ins-det cs.LG hep-ex | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The 75,848 lead tungstate crystals in CMS experiment at the CERN Large Hadron
Collider are used to measure the energy of electrons and photons produced in
the proton-proton collisions. The optical transparency of the crystals degrades
slowly with radiation dose due to the beam-beam collisions. The transparency of
each crystal is monitored with a laser monitoring system that tracks changes in
the optical properties of the crystals due to radiation from the collision
products. Predicting the optical transparency of the crystals, both in the
short-term and in the long-term, is a critical task for the CMS experiment. We
describe here the public data release, following FAIR principles, of the
crystal monitoring data collected by the CMS Collaboration between 2016 and
2018. Besides describing the dataset and its access, the problems that can be
addressed with it are described, as well as an example solution based on a Long
Short-Term Memory neural network developed to predict future behavior of the
crystals.
| [
{
"version": "v1",
"created": "Thu, 23 Mar 2023 20:14:15 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Joshi",
"Bhargav",
""
],
[
"Li",
"Taihui",
""
],
[
"Liang",
"Buyun",
""
],
[
"Rusack",
"Roger",
""
],
[
"Sun",
"Ju",
""
]
] | TITLE: Predicting the Future of the CMS Detector: Crystal Radiation Damage and
Machine Learning at the LHC
ABSTRACT: The 75,848 lead tungstate crystals in CMS experiment at the CERN Large Hadron
Collider are used to measure the energy of electrons and photons produced in
the proton-proton collisions. The optical transparency of the crystals degrades
slowly with radiation dose due to the beam-beam collisions. The transparency of
each crystal is monitored with a laser monitoring system that tracks changes in
the optical properties of the crystals due to radiation from the collision
products. Predicting the optical transparency of the crystals, both in the
short-term and in the long-term, is a critical task for the CMS experiment. We
describe here the public data release, following FAIR principles, of the
crystal monitoring data collected by the CMS Collaboration between 2016 and
2018. Besides describing the dataset and its access, the problems that can be
addressed with it are described, as well as an example solution based on a Long
Short-Term Memory neural network developed to predict future behavior of the
crystals.
| no_new_dataset | 0.950273 |
2303.15311 | Nicholas Monath | Nicholas Monath, Manzil Zaheer, Kelsey Allen, Andrew McCallum | Improving Dual-Encoder Training through Dynamic Indexes for Negative
Mining | To appear at AISTATS 2023 | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | Dual encoder models are ubiquitous in modern classification and retrieval.
Crucial for training such dual encoders is an accurate estimation of gradients
from the partition function of the softmax over the large output space; this
requires finding negative targets that contribute most significantly ("hard
negatives"). Since dual encoder model parameters change during training, the
use of traditional static nearest neighbor indexes can be sub-optimal. These
static indexes (1) periodically require expensive re-building of the index,
which in turn requires (2) expensive re-encoding of all targets using updated
model parameters. This paper addresses both of these challenges. First, we
introduce an algorithm that uses a tree structure to approximate the softmax
with provable bounds and that dynamically maintains the tree. Second, we
approximate the effect of a gradient update on target encodings with an
efficient Nystrom low-rank approximation. In our empirical study on datasets
with over twenty million targets, our approach cuts error by half in relation
to oracle brute-force negative mining. Furthermore, our method surpasses prior
state-of-the-art while using 150x less accelerator memory.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 15:18:32 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Monath",
"Nicholas",
""
],
[
"Zaheer",
"Manzil",
""
],
[
"Allen",
"Kelsey",
""
],
[
"McCallum",
"Andrew",
""
]
] | TITLE: Improving Dual-Encoder Training through Dynamic Indexes for Negative
Mining
ABSTRACT: Dual encoder models are ubiquitous in modern classification and retrieval.
Crucial for training such dual encoders is an accurate estimation of gradients
from the partition function of the softmax over the large output space; this
requires finding negative targets that contribute most significantly ("hard
negatives"). Since dual encoder model parameters change during training, the
use of traditional static nearest neighbor indexes can be sub-optimal. These
static indexes (1) periodically require expensive re-building of the index,
which in turn requires (2) expensive re-encoding of all targets using updated
model parameters. This paper addresses both of these challenges. First, we
introduce an algorithm that uses a tree structure to approximate the softmax
with provable bounds and that dynamically maintains the tree. Second, we
approximate the effect of a gradient update on target encodings with an
efficient Nystrom low-rank approximation. In our empirical study on datasets
with over twenty million targets, our approach cuts error by half in relation
to oracle brute-force negative mining. Furthermore, our method surpasses prior
state-of-the-art while using 150x less accelerator memory.
| no_new_dataset | 0.946892 |
2303.15331 | Arnaud Klipfel | Arnaud Klipfel, Nitish Sontakke, Ren Liu, Sehoon Ha | Learning a Single Policy for Diverse Behaviors on a Quadrupedal Robot
using Scalable Motion Imitation | null | null | null | null | cs.RO | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Learning various motor skills for quadrupedal robots is a challenging problem
that requires careful design of task-specific mathematical models or reward
descriptions. In this work, we propose to learn a single capable policy using
deep reinforcement learning by imitating a large number of reference motions,
including walking, turning, pacing, jumping, sitting, and lying. On top of the
existing motion imitation framework, we first carefully design the observation
space, the action space, and the reward function to improve the scalability of
the learning as well as the robustness of the final policy. In addition, we
adopt a novel adaptive motion sampling (AMS) method, which maintains a balance
between successful and unsuccessful behaviors. This technique allows the
learning algorithm to focus on challenging motor skills and avoid catastrophic
forgetting. We demonstrate that the learned policy can exhibit diverse
behaviors in simulation by successfully tracking both the training dataset and
out-of-distribution trajectories. We also validate the importance of the
proposed learning formulation and the adaptive motion sampling scheme by
conducting experiments.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 15:32:41 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Klipfel",
"Arnaud",
""
],
[
"Sontakke",
"Nitish",
""
],
[
"Liu",
"Ren",
""
],
[
"Ha",
"Sehoon",
""
]
] | TITLE: Learning a Single Policy for Diverse Behaviors on a Quadrupedal Robot
using Scalable Motion Imitation
ABSTRACT: Learning various motor skills for quadrupedal robots is a challenging problem
that requires careful design of task-specific mathematical models or reward
descriptions. In this work, we propose to learn a single capable policy using
deep reinforcement learning by imitating a large number of reference motions,
including walking, turning, pacing, jumping, sitting, and lying. On top of the
existing motion imitation framework, we first carefully design the observation
space, the action space, and the reward function to improve the scalability of
the learning as well as the robustness of the final policy. In addition, we
adopt a novel adaptive motion sampling (AMS) method, which maintains a balance
between successful and unsuccessful behaviors. This technique allows the
learning algorithm to focus on challenging motor skills and avoid catastrophic
forgetting. We demonstrate that the learned policy can exhibit diverse
behaviors in simulation by successfully tracking both the training dataset and
out-of-distribution trajectories. We also validate the importance of the
proposed learning formulation and the adaptive motion sampling scheme by
conducting experiments.
| no_new_dataset | 0.942188 |
2303.15342 | Micha{\l} Zaj\k{a}c | Micha{\l} Zaj\k{a}c, Kamil Deja, Anna Kuzina, Jakub M. Tomczak, Tomasz
Trzci\'nski, Florian Shkurti, Piotr Mi{\l}o\'s | Exploring Continual Learning of Diffusion Models | null | null | null | null | cs.LG cs.AI cs.CV stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Diffusion models have achieved remarkable success in generating high-quality
images thanks to their novel training procedures applied to unprecedented
amounts of data. However, training a diffusion model from scratch is
computationally expensive. This highlights the need to investigate the
possibility of training these models iteratively, reusing computation while the
data distribution changes. In this study, we take the first step in this
direction and evaluate the continual learning (CL) properties of diffusion
models. We begin by benchmarking the most common CL methods applied to
Denoising Diffusion Probabilistic Models (DDPMs), where we note the strong
performance of the experience replay with the reduced rehearsal coefficient.
Furthermore, we provide insights into the dynamics of forgetting, which exhibit
diverse behavior across diffusion timesteps. We also uncover certain pitfalls
of using the bits-per-dimension metric for evaluating CL.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 15:52:14 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Zając",
"Michał",
""
],
[
"Deja",
"Kamil",
""
],
[
"Kuzina",
"Anna",
""
],
[
"Tomczak",
"Jakub M.",
""
],
[
"Trzciński",
"Tomasz",
""
],
[
"Shkurti",
"Florian",
""
],
[
"Miłoś",
"Piotr",
""
]
] | TITLE: Exploring Continual Learning of Diffusion Models
ABSTRACT: Diffusion models have achieved remarkable success in generating high-quality
images thanks to their novel training procedures applied to unprecedented
amounts of data. However, training a diffusion model from scratch is
computationally expensive. This highlights the need to investigate the
possibility of training these models iteratively, reusing computation while the
data distribution changes. In this study, we take the first step in this
direction and evaluate the continual learning (CL) properties of diffusion
models. We begin by benchmarking the most common CL methods applied to
Denoising Diffusion Probabilistic Models (DDPMs), where we note the strong
performance of the experience replay with the reduced rehearsal coefficient.
Furthermore, we provide insights into the dynamics of forgetting, which exhibit
diverse behavior across diffusion timesteps. We also uncover certain pitfalls
of using the bits-per-dimension metric for evaluating CL.
| no_new_dataset | 0.950227 |
2303.15377 | Wei Lin | Kunyang Sun, Wei Lin, Haoqin Shi, Zhengming Zhang, Yongming Huang,
Horst Bischof | AIR-DA: Adversarial Image Reconstruction for Unsupervised Domain
Adaptive Object Detection | Accepted at IEEE Robotics and Automation Letters 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Unsupervised domain adaptive object detection is a challenging vision task
where object detectors are adapted from a label-rich source domain to an
unlabeled target domain. Recent advances prove the efficacy of the adversarial
based domain alignment where the adversarial training between the feature
extractor and domain discriminator results in domain-invariance in the feature
space. However, due to the domain shift, domain discrimination, especially on
low-level features, is an easy task. This results in an imbalance of the
adversarial training between the domain discriminator and the feature
extractor. In this work, we achieve a better domain alignment by introducing an
auxiliary regularization task to improve the training balance. Specifically, we
propose Adversarial Image Reconstruction (AIR) as the regularizer to facilitate
the adversarial training of the feature extractor. We further design a
multi-level feature alignment module to enhance the adaptation performance. Our
evaluations across several datasets of challenging domain shifts demonstrate
that the proposed method outperforms all previous methods, of both one- and
two-stage, in most settings.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 16:51:51 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Sun",
"Kunyang",
""
],
[
"Lin",
"Wei",
""
],
[
"Shi",
"Haoqin",
""
],
[
"Zhang",
"Zhengming",
""
],
[
"Huang",
"Yongming",
""
],
[
"Bischof",
"Horst",
""
]
] | TITLE: AIR-DA: Adversarial Image Reconstruction for Unsupervised Domain
Adaptive Object Detection
ABSTRACT: Unsupervised domain adaptive object detection is a challenging vision task
where object detectors are adapted from a label-rich source domain to an
unlabeled target domain. Recent advances prove the efficacy of the adversarial
based domain alignment where the adversarial training between the feature
extractor and domain discriminator results in domain-invariance in the feature
space. However, due to the domain shift, domain discrimination, especially on
low-level features, is an easy task. This results in an imbalance of the
adversarial training between the domain discriminator and the feature
extractor. In this work, we achieve a better domain alignment by introducing an
auxiliary regularization task to improve the training balance. Specifically, we
propose Adversarial Image Reconstruction (AIR) as the regularizer to facilitate
the adversarial training of the feature extractor. We further design a
multi-level feature alignment module to enhance the adaptation performance. Our
evaluations across several datasets of challenging domain shifts demonstrate
that the proposed method outperforms all previous methods, of both one- and
two-stage, in most settings.
| no_new_dataset | 0.946547 |
2303.15380 | Chen Guo | Yifei Yin, Chen Guo, Manuel Kaufmann, Juan Jose Zarate, Jie Song,
Otmar Hilliges | Hi4D: 4D Instance Segmentation of Close Human Interaction | Project page: https://yifeiyin04.github.io/Hi4D/ | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose Hi4D, a method and dataset for the automatic analysis of
physically close human-human interaction under prolonged contact. Robustly
disentangling several in-contact subjects is a challenging task due to
occlusions and complex shapes. Hence, existing multi-view systems typically
fuse 3D surfaces of close subjects into a single, connected mesh. To address
this issue we leverage i) individually fitted neural implicit avatars; ii) an
alternating optimization scheme that refines pose and surface through periods
of close proximity; and iii) thus segment the fused raw scans into individual
instances. From these instances we compile Hi4D dataset of 4D textured scans of
20 subject pairs, 100 sequences, and a total of more than 11K frames. Hi4D
contains rich interaction-centric annotations in 2D and 3D alongside accurately
registered parametric body models. We define varied human pose and shape
estimation tasks on this dataset and provide results from state-of-the-art
methods on these benchmarks.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 16:53:09 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Yin",
"Yifei",
""
],
[
"Guo",
"Chen",
""
],
[
"Kaufmann",
"Manuel",
""
],
[
"Zarate",
"Juan Jose",
""
],
[
"Song",
"Jie",
""
],
[
"Hilliges",
"Otmar",
""
]
] | TITLE: Hi4D: 4D Instance Segmentation of Close Human Interaction
ABSTRACT: We propose Hi4D, a method and dataset for the automatic analysis of
physically close human-human interaction under prolonged contact. Robustly
disentangling several in-contact subjects is a challenging task due to
occlusions and complex shapes. Hence, existing multi-view systems typically
fuse 3D surfaces of close subjects into a single, connected mesh. To address
this issue we leverage i) individually fitted neural implicit avatars; ii) an
alternating optimization scheme that refines pose and surface through periods
of close proximity; and iii) thus segment the fused raw scans into individual
instances. From these instances we compile Hi4D dataset of 4D textured scans of
20 subject pairs, 100 sequences, and a total of more than 11K frames. Hi4D
contains rich interaction-centric annotations in 2D and 3D alongside accurately
registered parametric body models. We define varied human pose and shape
estimation tasks on this dataset and provide results from state-of-the-art
methods on these benchmarks.
| new_dataset | 0.959001 |
2303.15381 | Michael Regan | Michael Regan and Jena D. Hwang and Keisuke Sakaguchi and James
Pustejovsky | Causal schema induction for knowledge discovery | 8 pages, appendix | null | null | null | cs.CL | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Making sense of familiar yet new situations typically involves making
generalizations about causal schemas, stories that help humans reason about
event sequences. Reasoning about events includes identifying cause and effect
relations shared across event instances, a process we refer to as causal schema
induction. Statistical schema induction systems may leverage structural
knowledge encoded in discourse or the causal graphs associated with event
meaning, however resources to study such causal structure are few in number and
limited in size. In this work, we investigate how to apply schema induction
models to the task of knowledge discovery for enhanced search of
English-language news texts. To tackle the problem of data scarcity, we present
Torquestra, a manually curated dataset of text-graph-schema units integrating
temporal, event, and causal structures. We benchmark our dataset on three
knowledge discovery tasks, building and evaluating models for each. Results
show that systems that harness causal structure are effective at identifying
texts sharing similar causal meaning components rather than relying on lexical
cues alone. We make our dataset and models available for research purposes.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 16:55:49 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Regan",
"Michael",
""
],
[
"Hwang",
"Jena D.",
""
],
[
"Sakaguchi",
"Keisuke",
""
],
[
"Pustejovsky",
"James",
""
]
] | TITLE: Causal schema induction for knowledge discovery
ABSTRACT: Making sense of familiar yet new situations typically involves making
generalizations about causal schemas, stories that help humans reason about
event sequences. Reasoning about events includes identifying cause and effect
relations shared across event instances, a process we refer to as causal schema
induction. Statistical schema induction systems may leverage structural
knowledge encoded in discourse or the causal graphs associated with event
meaning, however resources to study such causal structure are few in number and
limited in size. In this work, we investigate how to apply schema induction
models to the task of knowledge discovery for enhanced search of
English-language news texts. To tackle the problem of data scarcity, we present
Torquestra, a manually curated dataset of text-graph-schema units integrating
temporal, event, and causal structures. We benchmark our dataset on three
knowledge discovery tasks, building and evaluating models for each. Results
show that systems that harness causal structure are effective at identifying
texts sharing similar causal meaning components rather than relying on lexical
cues alone. We make our dataset and models available for research purposes.
| new_dataset | 0.956877 |
2303.15390 | Chittesh Thavamani | Chittesh Thavamani, Mengtian Li, Francesco Ferroni, Deva Ramanan | Learning to Zoom and Unzoom | CVPR 2023. Code and additional visuals available at
https://tchittesh.github.io/lzu/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Many perception systems in mobile computing, autonomous navigation, and AR/VR
face strict compute constraints that are particularly challenging for
high-resolution input images. Previous works propose nonuniform downsamplers
that "learn to zoom" on salient image regions, reducing compute while retaining
task-relevant image information. However, for tasks with spatial labels (such
as 2D/3D object detection and semantic segmentation), such distortions may harm
performance. In this work (LZU), we "learn to zoom" in on the input image,
compute spatial features, and then "unzoom" to revert any deformations. To
enable efficient and differentiable unzooming, we approximate the zooming warp
with a piecewise bilinear mapping that is invertible. LZU can be applied to any
task with 2D spatial input and any model with 2D spatial features, and we
demonstrate this versatility by evaluating on a variety of tasks and datasets:
object detection on Argoverse-HD, semantic segmentation on Cityscapes, and
monocular 3D object detection on nuScenes. Interestingly, we observe boosts in
performance even when high-resolution sensor data is unavailable, implying that
LZU can be used to "learn to upsample" as well.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:03:30 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Thavamani",
"Chittesh",
""
],
[
"Li",
"Mengtian",
""
],
[
"Ferroni",
"Francesco",
""
],
[
"Ramanan",
"Deva",
""
]
] | TITLE: Learning to Zoom and Unzoom
ABSTRACT: Many perception systems in mobile computing, autonomous navigation, and AR/VR
face strict compute constraints that are particularly challenging for
high-resolution input images. Previous works propose nonuniform downsamplers
that "learn to zoom" on salient image regions, reducing compute while retaining
task-relevant image information. However, for tasks with spatial labels (such
as 2D/3D object detection and semantic segmentation), such distortions may harm
performance. In this work (LZU), we "learn to zoom" in on the input image,
compute spatial features, and then "unzoom" to revert any deformations. To
enable efficient and differentiable unzooming, we approximate the zooming warp
with a piecewise bilinear mapping that is invertible. LZU can be applied to any
task with 2D spatial input and any model with 2D spatial features, and we
demonstrate this versatility by evaluating on a variety of tasks and datasets:
object detection on Argoverse-HD, semantic segmentation on Cityscapes, and
monocular 3D object detection on nuScenes. Interestingly, we observe boosts in
performance even when high-resolution sensor data is unavailable, implying that
LZU can be used to "learn to upsample" as well.
| no_new_dataset | 0.942876 |
2303.15410 | Sohyun Lee | Sohyun Lee, Jaesung Rim, Boseung Jeong, Geonu Kim, Byungju Woo,
Haechan Lee, Sunghyun Cho, Suha Kwak | Human Pose Estimation in Extremely Low-Light Conditions | Accepted to CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We study human pose estimation in extremely low-light images. This task is
challenging due to the difficulty of collecting real low-light images with
accurate labels, and severely corrupted inputs that degrade prediction quality
significantly. To address the first issue, we develop a dedicated camera system
and build a new dataset of real low-light images with accurate pose labels.
Thanks to our camera system, each low-light image in our dataset is coupled
with an aligned well-lit image, which enables accurate pose labeling and is
used as privileged information during training. We also propose a new model and
a new training strategy that fully exploit the privileged information to learn
representation insensitive to lighting conditions. Our method demonstrates
outstanding performance on real extremely low light images, and extensive
analyses validate that both of our model and dataset contribute to the success.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:28:25 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Lee",
"Sohyun",
""
],
[
"Rim",
"Jaesung",
""
],
[
"Jeong",
"Boseung",
""
],
[
"Kim",
"Geonu",
""
],
[
"Woo",
"Byungju",
""
],
[
"Lee",
"Haechan",
""
],
[
"Cho",
"Sunghyun",
""
],
[
"Kwak",
"Suha",
""
]
] | TITLE: Human Pose Estimation in Extremely Low-Light Conditions
ABSTRACT: We study human pose estimation in extremely low-light images. This task is
challenging due to the difficulty of collecting real low-light images with
accurate labels, and severely corrupted inputs that degrade prediction quality
significantly. To address the first issue, we develop a dedicated camera system
and build a new dataset of real low-light images with accurate pose labels.
Thanks to our camera system, each low-light image in our dataset is coupled
with an aligned well-lit image, which enables accurate pose labeling and is
used as privileged information during training. We also propose a new model and
a new training strategy that fully exploit the privileged information to learn
representation insensitive to lighting conditions. Our method demonstrates
outstanding performance on real extremely low light images, and extensive
analyses validate that both of our model and dataset contribute to the success.
| new_dataset | 0.953013 |
2303.15417 | Jaeha Kim | Yeonguk Oh, JoonKyu Park, Jaeha Kim, Gyeongsik Moon, Kyoung Mu Lee | Recovering 3D Hand Mesh Sequence from a Single Blurry Image: A New
Dataset and Temporal Unfolding | Accepted at CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Hands, one of the most dynamic parts of our body, suffer from blur due to
their active movements. However, previous 3D hand mesh recovery methods have
mainly focused on sharp hand images rather than considering blur due to the
absence of datasets providing blurry hand images. We first present a novel
dataset BlurHand, which contains blurry hand images with 3D groundtruths. The
BlurHand is constructed by synthesizing motion blur from sequential sharp hand
images, imitating realistic and natural motion blurs. In addition to the new
dataset, we propose BlurHandNet, a baseline network for accurate 3D hand mesh
recovery from a blurry hand image. Our BlurHandNet unfolds a blurry input image
to a 3D hand mesh sequence to utilize temporal information in the blurry input
image, while previous works output a static single hand mesh. We demonstrate
the usefulness of BlurHand for the 3D hand mesh recovery from blurry images in
our experiments. The proposed BlurHandNet produces much more robust results on
blurry images while generalizing well to in-the-wild images. The training codes
and BlurHand dataset are available at
https://github.com/JaehaKim97/BlurHand_RELEASE.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:40:29 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Oh",
"Yeonguk",
""
],
[
"Park",
"JoonKyu",
""
],
[
"Kim",
"Jaeha",
""
],
[
"Moon",
"Gyeongsik",
""
],
[
"Lee",
"Kyoung Mu",
""
]
] | TITLE: Recovering 3D Hand Mesh Sequence from a Single Blurry Image: A New
Dataset and Temporal Unfolding
ABSTRACT: Hands, one of the most dynamic parts of our body, suffer from blur due to
their active movements. However, previous 3D hand mesh recovery methods have
mainly focused on sharp hand images rather than considering blur due to the
absence of datasets providing blurry hand images. We first present a novel
dataset BlurHand, which contains blurry hand images with 3D groundtruths. The
BlurHand is constructed by synthesizing motion blur from sequential sharp hand
images, imitating realistic and natural motion blurs. In addition to the new
dataset, we propose BlurHandNet, a baseline network for accurate 3D hand mesh
recovery from a blurry hand image. Our BlurHandNet unfolds a blurry input image
to a 3D hand mesh sequence to utilize temporal information in the blurry input
image, while previous works output a static single hand mesh. We demonstrate
the usefulness of BlurHand for the 3D hand mesh recovery from blurry images in
our experiments. The proposed BlurHandNet produces much more robust results on
blurry images while generalizing well to in-the-wild images. The training codes
and BlurHand dataset are available at
https://github.com/JaehaKim97/BlurHand_RELEASE.
| new_dataset | 0.958886 |
2303.15437 | Anurag Ranjan | Anurag Ranjan, Kwang Moo Yi, Jen-Hao Rick Chang, Oncel Tuzel | FaceLit: Neural 3D Relightable Faces | CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We propose a generative framework, FaceLit, capable of generating a 3D face
that can be rendered at various user-defined lighting conditions and views,
learned purely from 2D images in-the-wild without any manual annotation. Unlike
existing works that require careful capture setup or human labor, we rely on
off-the-shelf pose and illumination estimators. With these estimates, we
incorporate the Phong reflectance model in the neural volume rendering
framework. Our model learns to generate shape and material properties of a face
such that, when rendered according to the natural statistics of pose and
illumination, produces photorealistic face images with multiview 3D and
illumination consistency. Our method enables photorealistic generation of faces
with explicit illumination and view controls on multiple datasets - FFHQ,
MetFaces and CelebA-HQ. We show state-of-the-art photorealism among 3D aware
GANs on FFHQ dataset achieving an FID score of 3.5.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:59:10 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Ranjan",
"Anurag",
""
],
[
"Yi",
"Kwang Moo",
""
],
[
"Chang",
"Jen-Hao Rick",
""
],
[
"Tuzel",
"Oncel",
""
]
] | TITLE: FaceLit: Neural 3D Relightable Faces
ABSTRACT: We propose a generative framework, FaceLit, capable of generating a 3D face
that can be rendered at various user-defined lighting conditions and views,
learned purely from 2D images in-the-wild without any manual annotation. Unlike
existing works that require careful capture setup or human labor, we rely on
off-the-shelf pose and illumination estimators. With these estimates, we
incorporate the Phong reflectance model in the neural volume rendering
framework. Our model learns to generate shape and material properties of a face
such that, when rendered according to the natural statistics of pose and
illumination, produces photorealistic face images with multiview 3D and
illumination consistency. Our method enables photorealistic generation of faces
with explicit illumination and view controls on multiple datasets - FFHQ,
MetFaces and CelebA-HQ. We show state-of-the-art photorealism among 3D aware
GANs on FFHQ dataset achieving an FID score of 3.5.
| no_new_dataset | 0.951594 |
2303.15440 | Jiahui Lei | Jiahui Lei and Congyue Deng and Karl Schmeckpeper and Leonidas Guibas
and Kostas Daniilidis | EFEM: Equivariant Neural Field Expectation Maximization for 3D Object
Segmentation Without Scene Supervision | Accepted by CVPR2023, project page
https://www.cis.upenn.edu/~leijh/projects/efem | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We introduce Equivariant Neural Field Expectation Maximization (EFEM), a
simple, effective, and robust geometric algorithm that can segment objects in
3D scenes without annotations or training on scenes. We achieve such
unsupervised segmentation by exploiting single object shape priors. We make two
novel steps in that direction. First, we introduce equivariant shape
representations to this problem to eliminate the complexity induced by the
variation in object configuration. Second, we propose a novel EM algorithm that
can iteratively refine segmentation masks using the equivariant shape prior. We
collect a novel real dataset Chairs and Mugs that contains various object
configurations and novel scenes in order to verify the effectiveness and
robustness of our method. Experimental results demonstrate that our method
achieves consistent and robust performance across different scenes where the
(weakly) supervised methods may fail. Code and data available at
https://www.cis.upenn.edu/~leijh/projects/efem
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:59:29 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Lei",
"Jiahui",
""
],
[
"Deng",
"Congyue",
""
],
[
"Schmeckpeper",
"Karl",
""
],
[
"Guibas",
"Leonidas",
""
],
[
"Daniilidis",
"Kostas",
""
]
] | TITLE: EFEM: Equivariant Neural Field Expectation Maximization for 3D Object
Segmentation Without Scene Supervision
ABSTRACT: We introduce Equivariant Neural Field Expectation Maximization (EFEM), a
simple, effective, and robust geometric algorithm that can segment objects in
3D scenes without annotations or training on scenes. We achieve such
unsupervised segmentation by exploiting single object shape priors. We make two
novel steps in that direction. First, we introduce equivariant shape
representations to this problem to eliminate the complexity induced by the
variation in object configuration. Second, we propose a novel EM algorithm that
can iteratively refine segmentation masks using the equivariant shape prior. We
collect a novel real dataset Chairs and Mugs that contains various object
configurations and novel scenes in order to verify the effectiveness and
robustness of our method. Experimental results demonstrate that our method
achieves consistent and robust performance across different scenes where the
(weakly) supervised methods may fail. Code and data available at
https://www.cis.upenn.edu/~leijh/projects/efem
| new_dataset | 0.95635 |
2303.15443 | Tarun Kalluri | Tarun Kalluri, Wangdong Xu, Manmohan Chandraker | GeoNet: Benchmarking Unsupervised Adaptation across Geographies | CVPR 2023 Camera Ready. Project Page:
https://tarun005.github.io/GeoNet | null | null | null | cs.CV cs.AI cs.LG | http://creativecommons.org/licenses/by/4.0/ | In recent years, several efforts have been aimed at improving the robustness
of vision models to domains and environments unseen during training. An
important practical problem pertains to models deployed in a new geography that
is under-represented in the training dataset, posing a direct challenge to fair
and inclusive computer vision. In this paper, we study the problem of
geographic robustness and make three main contributions. First, we introduce a
large-scale dataset GeoNet for geographic adaptation containing benchmarks
across diverse tasks like scene recognition (GeoPlaces), image classification
(GeoImNet) and universal adaptation (GeoUniDA). Second, we investigate the
nature of distribution shifts typical to the problem of geographic adaptation
and hypothesize that the major source of domain shifts arise from significant
variations in scene context (context shift), object design (design shift) and
label distribution (prior shift) across geographies. Third, we conduct an
extensive evaluation of several state-of-the-art unsupervised domain adaptation
algorithms and architectures on GeoNet, showing that they do not suffice for
geographical adaptation, and that large-scale pre-training using large vision
models also does not lead to geographic robustness. Our dataset is publicly
available at https://tarun005.github.io/GeoNet.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:59:34 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Kalluri",
"Tarun",
""
],
[
"Xu",
"Wangdong",
""
],
[
"Chandraker",
"Manmohan",
""
]
] | TITLE: GeoNet: Benchmarking Unsupervised Adaptation across Geographies
ABSTRACT: In recent years, several efforts have been aimed at improving the robustness
of vision models to domains and environments unseen during training. An
important practical problem pertains to models deployed in a new geography that
is under-represented in the training dataset, posing a direct challenge to fair
and inclusive computer vision. In this paper, we study the problem of
geographic robustness and make three main contributions. First, we introduce a
large-scale dataset GeoNet for geographic adaptation containing benchmarks
across diverse tasks like scene recognition (GeoPlaces), image classification
(GeoImNet) and universal adaptation (GeoUniDA). Second, we investigate the
nature of distribution shifts typical to the problem of geographic adaptation
and hypothesize that the major source of domain shifts arise from significant
variations in scene context (context shift), object design (design shift) and
label distribution (prior shift) across geographies. Third, we conduct an
extensive evaluation of several state-of-the-art unsupervised domain adaptation
algorithms and architectures on GeoNet, showing that they do not suffice for
geographical adaptation, and that large-scale pre-training using large vision
models also does not lead to geographic robustness. Our dataset is publicly
available at https://tarun005.github.io/GeoNet.
| new_dataset | 0.962497 |
2303.15444 | Vladislav Golyanik | Matteo Farina and Luca Magri and Willi Menapace and Elisa Ricci and
Vladislav Golyanik and Federica Arrigoni | Quantum Multi-Model Fitting | In Computer Vision and Pattern Recognition (CVPR) 2023; Highlight | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Geometric model fitting is a challenging but fundamental computer vision
problem. Recently, quantum optimization has been shown to enhance robust
fitting for the case of a single model, while leaving the question of
multi-model fitting open. In response to this challenge, this paper shows that
the latter case can significantly benefit from quantum hardware and proposes
the first quantum approach to multi-model fitting (MMF). We formulate MMF as a
problem that can be efficiently sampled by modern adiabatic quantum computers
without the relaxation of the objective function. We also propose an iterative
and decomposed version of our method, which supports real-world-sized problems.
The experimental evaluation demonstrates promising results on a variety of
datasets. The source code is available at:
https://github.com/FarinaMatteo/qmmf.
| [
{
"version": "v1",
"created": "Mon, 27 Mar 2023 17:59:54 GMT"
}
] | 2023-03-28T00:00:00 | [
[
"Farina",
"Matteo",
""
],
[
"Magri",
"Luca",
""
],
[
"Menapace",
"Willi",
""
],
[
"Ricci",
"Elisa",
""
],
[
"Golyanik",
"Vladislav",
""
],
[
"Arrigoni",
"Federica",
""
]
] | TITLE: Quantum Multi-Model Fitting
ABSTRACT: Geometric model fitting is a challenging but fundamental computer vision
problem. Recently, quantum optimization has been shown to enhance robust
fitting for the case of a single model, while leaving the question of
multi-model fitting open. In response to this challenge, this paper shows that
the latter case can significantly benefit from quantum hardware and proposes
the first quantum approach to multi-model fitting (MMF). We formulate MMF as a
problem that can be efficiently sampled by modern adiabatic quantum computers
without the relaxation of the objective function. We also propose an iterative
and decomposed version of our method, which supports real-world-sized problems.
The experimental evaluation demonstrates promising results on a variety of
datasets. The source code is available at:
https://github.com/FarinaMatteo/qmmf.
| no_new_dataset | 0.949295 |
2011.13704 | Jakob Drefs | Enrico Guiraud, Jakob Drefs, J\"org L\"ucke | Direct Evolutionary Optimization of Variational Autoencoders With Binary
Latents | null | null | 10.1007/978-3-031-26409-2_22 | null | stat.ML cs.LG | http://creativecommons.org/licenses/by/4.0/ | Discrete latent variables are considered important for real world data, which
has motivated research on Variational Autoencoders (VAEs) with discrete
latents. However, standard VAE training is not possible in this case, which has
motivated different strategies to manipulate discrete distributions in order to
train discrete VAEs similarly to conventional ones. Here we ask if it is also
possible to keep the discrete nature of the latents fully intact by applying a
direct discrete optimization for the encoding model. The approach is
consequently strongly diverting from standard VAE-training by sidestepping
sampling approximation, reparameterization trick and amortization. Discrete
optimization is realized in a variational setting using truncated posteriors in
conjunction with evolutionary algorithms. For VAEs with binary latents, we (A)
show how such a discrete variational method ties into gradient ascent for
network weights, and (B) how the decoder is used to select latent states for
training. Conventional amortized training is more efficient and applicable to
large neural networks. However, using smaller networks, we here find direct
discrete optimization to be efficiently scalable to hundreds of latents. More
importantly, we find the effectiveness of direct optimization to be highly
competitive in `zero-shot' learning. In contrast to large supervised networks,
the here investigated VAEs can, e.g., denoise a single image without previous
training on clean data and/or training on large image datasets. More generally,
the studied approach shows that training of VAEs is indeed possible without
sampling-based approximation and reparameterization, which may be interesting
for the analysis of VAE-training in general. For `zero-shot' settings a direct
optimization, furthermore, makes VAEs competitive where they have previously
been outperformed by non-generative approaches.
| [
{
"version": "v1",
"created": "Fri, 27 Nov 2020 12:42:12 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 13:14:14 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Guiraud",
"Enrico",
""
],
[
"Drefs",
"Jakob",
""
],
[
"Lücke",
"Jörg",
""
]
] | TITLE: Direct Evolutionary Optimization of Variational Autoencoders With Binary
Latents
ABSTRACT: Discrete latent variables are considered important for real world data, which
has motivated research on Variational Autoencoders (VAEs) with discrete
latents. However, standard VAE training is not possible in this case, which has
motivated different strategies to manipulate discrete distributions in order to
train discrete VAEs similarly to conventional ones. Here we ask if it is also
possible to keep the discrete nature of the latents fully intact by applying a
direct discrete optimization for the encoding model. The approach is
consequently strongly diverting from standard VAE-training by sidestepping
sampling approximation, reparameterization trick and amortization. Discrete
optimization is realized in a variational setting using truncated posteriors in
conjunction with evolutionary algorithms. For VAEs with binary latents, we (A)
show how such a discrete variational method ties into gradient ascent for
network weights, and (B) how the decoder is used to select latent states for
training. Conventional amortized training is more efficient and applicable to
large neural networks. However, using smaller networks, we here find direct
discrete optimization to be efficiently scalable to hundreds of latents. More
importantly, we find the effectiveness of direct optimization to be highly
competitive in `zero-shot' learning. In contrast to large supervised networks,
the here investigated VAEs can, e.g., denoise a single image without previous
training on clean data and/or training on large image datasets. More generally,
the studied approach shows that training of VAEs is indeed possible without
sampling-based approximation and reparameterization, which may be interesting
for the analysis of VAE-training in general. For `zero-shot' settings a direct
optimization, furthermore, makes VAEs competitive where they have previously
been outperformed by non-generative approaches.
| no_new_dataset | 0.943504 |
2107.04466 | Jonathan Siegel | Jonathan W. Siegel, Qingguo Hong, Xianlin Jin, Wenrui Hao and Jinchao
Xu | Greedy Training Algorithms for Neural Networks and Applications to PDEs | has been merged with arXiv:2104.02903 | null | null | null | math.NA cs.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Recently, neural networks have been widely applied for solving partial
differential equations (PDEs). Although such methods have been proven
remarkably successful on practical engineering problems, they have not been
shown, theoretically or empirically, to converge to the underlying PDE solution
with arbitrarily high accuracy. The primary difficulty lies in solving the
highly non-convex optimization problems resulting from the neural network
discretization, which are difficult to treat both theoretically and
practically. It is our goal in this work to take a step toward remedying this.
For this purpose, we develop a novel greedy training algorithm for shallow
neural networks. Our method is applicable to both the variational formulation
of the PDE and also to the residual minimization formulation pioneered by
physics informed neural networks (PINNs). We analyze the method and obtain a
priori error bounds when solving PDEs from the function class defined by
shallow networks, which rigorously establishes the convergence of the method as
the network size increases. Finally, we test the algorithm on several benchmark
examples, including high dimensional PDEs, to confirm the theoretical
convergence rate. Although the method is expensive relative to traditional
approaches such as finite element methods, we view this work as a proof of
concept for neural network-based methods, which shows that numerical methods
based upon neural networks can be shown to rigorously converge.
| [
{
"version": "v1",
"created": "Fri, 9 Jul 2021 14:32:02 GMT"
},
{
"version": "v2",
"created": "Thu, 14 Jul 2022 18:53:11 GMT"
},
{
"version": "v3",
"created": "Fri, 23 Dec 2022 19:05:47 GMT"
},
{
"version": "v4",
"created": "Fri, 24 Mar 2023 17:51:27 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Siegel",
"Jonathan W.",
""
],
[
"Hong",
"Qingguo",
""
],
[
"Jin",
"Xianlin",
""
],
[
"Hao",
"Wenrui",
""
],
[
"Xu",
"Jinchao",
""
]
] | TITLE: Greedy Training Algorithms for Neural Networks and Applications to PDEs
ABSTRACT: Recently, neural networks have been widely applied for solving partial
differential equations (PDEs). Although such methods have been proven
remarkably successful on practical engineering problems, they have not been
shown, theoretically or empirically, to converge to the underlying PDE solution
with arbitrarily high accuracy. The primary difficulty lies in solving the
highly non-convex optimization problems resulting from the neural network
discretization, which are difficult to treat both theoretically and
practically. It is our goal in this work to take a step toward remedying this.
For this purpose, we develop a novel greedy training algorithm for shallow
neural networks. Our method is applicable to both the variational formulation
of the PDE and also to the residual minimization formulation pioneered by
physics informed neural networks (PINNs). We analyze the method and obtain a
priori error bounds when solving PDEs from the function class defined by
shallow networks, which rigorously establishes the convergence of the method as
the network size increases. Finally, we test the algorithm on several benchmark
examples, including high dimensional PDEs, to confirm the theoretical
convergence rate. Although the method is expensive relative to traditional
approaches such as finite element methods, we view this work as a proof of
concept for neural network-based methods, which shows that numerical methods
based upon neural networks can be shown to rigorously converge.
| no_new_dataset | 0.947186 |
2110.06650 | Andreas Triantafyllopoulos | Andreas Triantafyllopoulos, Uwe Reichel, Shuo Liu, Stephan Huber,
Florian Eyben, Bj\"orn W. Schuller | Multistage linguistic conditioning of convolutional layers for speech
emotion recognition | null | Frontiers in Computer Science, Volume 5, 2023 | 10.3389/fcomp.2023.1072479 | null | cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | In this contribution, we investigate the effectiveness of deep fusion of text
and audio features for categorical and dimensional speech emotion recognition
(SER). We propose a novel, multistage fusion method where the two information
streams are integrated in several layers of a deep neural network (DNN), and
contrast it with a single-stage one where the streams are merged in a single
point. Both methods depend on extracting summary linguistic embeddings from a
pre-trained BERT model, and conditioning one or more intermediate
representations of a convolutional model operating on log-Mel spectrograms.
Experiments on the MSP-Podcast and IEMOCAP datasets demonstrate that the two
fusion methods clearly outperform a shallow (late) fusion baseline and their
unimodal constituents, both in terms of quantitative performance and
qualitative behaviour. Overall, our multistage fusion shows better quantitative
performance, surpassing alternatives on most of our evaluations. This
illustrates the potential of multistage fusion in better assimilating text and
audio information.
| [
{
"version": "v1",
"created": "Wed, 13 Oct 2021 11:28:04 GMT"
},
{
"version": "v2",
"created": "Wed, 4 May 2022 13:33:20 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Triantafyllopoulos",
"Andreas",
""
],
[
"Reichel",
"Uwe",
""
],
[
"Liu",
"Shuo",
""
],
[
"Huber",
"Stephan",
""
],
[
"Eyben",
"Florian",
""
],
[
"Schuller",
"Björn W.",
""
]
] | TITLE: Multistage linguistic conditioning of convolutional layers for speech
emotion recognition
ABSTRACT: In this contribution, we investigate the effectiveness of deep fusion of text
and audio features for categorical and dimensional speech emotion recognition
(SER). We propose a novel, multistage fusion method where the two information
streams are integrated in several layers of a deep neural network (DNN), and
contrast it with a single-stage one where the streams are merged in a single
point. Both methods depend on extracting summary linguistic embeddings from a
pre-trained BERT model, and conditioning one or more intermediate
representations of a convolutional model operating on log-Mel spectrograms.
Experiments on the MSP-Podcast and IEMOCAP datasets demonstrate that the two
fusion methods clearly outperform a shallow (late) fusion baseline and their
unimodal constituents, both in terms of quantitative performance and
qualitative behaviour. Overall, our multistage fusion shows better quantitative
performance, surpassing alternatives on most of our evaluations. This
illustrates the potential of multistage fusion in better assimilating text and
audio information.
| no_new_dataset | 0.947381 |
2111.09543 | Pengcheng He | Pengcheng He, Jianfeng Gao and Weizhu Chen | DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with
Gradient-Disentangled Embedding Sharing | 16 pages, 10 tables, 2 Figures. The DeBERTaV3 model significantly
improves performance of the downstream NLU tasks over models with a similar
structure, e.g. DeBERTaV3 large achieves 91.37% average GLUE score which is
1.37% over DeBERTa large. XSmall has only 22M backbone parameters, but
significantly outperforms RoBERTa/XLNet-base. Paper is published as a
conference paper at ICLR 2023 | null | null | null | cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | This paper presents a new pre-trained language model, DeBERTaV3, which
improves the original DeBERTa model by replacing mask language modeling (MLM)
with replaced token detection (RTD), a more sample-efficient pre-training task.
Our analysis shows that vanilla embedding sharing in ELECTRA hurts training
efficiency and model performance. This is because the training losses of the
discriminator and the generator pull token embeddings in different directions,
creating the "tug-of-war" dynamics. We thus propose a new gradient-disentangled
embedding sharing method that avoids the tug-of-war dynamics, improving both
training efficiency and the quality of the pre-trained model. We have
pre-trained DeBERTaV3 using the same settings as DeBERTa to demonstrate its
exceptional performance on a wide range of downstream natural language
understanding (NLU) tasks. Taking the GLUE benchmark with eight tasks as an
example, the DeBERTaV3 Large model achieves a 91.37% average score, which is
1.37% over DeBERTa and 1.91% over ELECTRA, setting a new state-of-the-art
(SOTA) among the models with a similar structure. Furthermore, we have
pre-trained a multi-lingual model mDeBERTa and observed a larger improvement
over strong baselines compared to English models. For example, the mDeBERTa
Base achieves a 79.8% zero-shot cross-lingual accuracy on XNLI and a 3.6%
improvement over XLM-R Base, creating a new SOTA on this benchmark. We have
made our pre-trained models and inference code publicly available at
https://github.com/microsoft/DeBERTa.
| [
{
"version": "v1",
"created": "Thu, 18 Nov 2021 06:48:00 GMT"
},
{
"version": "v2",
"created": "Wed, 8 Dec 2021 22:07:23 GMT"
},
{
"version": "v3",
"created": "Tue, 21 Mar 2023 05:17:08 GMT"
},
{
"version": "v4",
"created": "Fri, 24 Mar 2023 09:17:17 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"He",
"Pengcheng",
""
],
[
"Gao",
"Jianfeng",
""
],
[
"Chen",
"Weizhu",
""
]
] | TITLE: DeBERTaV3: Improving DeBERTa using ELECTRA-Style Pre-Training with
Gradient-Disentangled Embedding Sharing
ABSTRACT: This paper presents a new pre-trained language model, DeBERTaV3, which
improves the original DeBERTa model by replacing mask language modeling (MLM)
with replaced token detection (RTD), a more sample-efficient pre-training task.
Our analysis shows that vanilla embedding sharing in ELECTRA hurts training
efficiency and model performance. This is because the training losses of the
discriminator and the generator pull token embeddings in different directions,
creating the "tug-of-war" dynamics. We thus propose a new gradient-disentangled
embedding sharing method that avoids the tug-of-war dynamics, improving both
training efficiency and the quality of the pre-trained model. We have
pre-trained DeBERTaV3 using the same settings as DeBERTa to demonstrate its
exceptional performance on a wide range of downstream natural language
understanding (NLU) tasks. Taking the GLUE benchmark with eight tasks as an
example, the DeBERTaV3 Large model achieves a 91.37% average score, which is
1.37% over DeBERTa and 1.91% over ELECTRA, setting a new state-of-the-art
(SOTA) among the models with a similar structure. Furthermore, we have
pre-trained a multi-lingual model mDeBERTa and observed a larger improvement
over strong baselines compared to English models. For example, the mDeBERTa
Base achieves a 79.8% zero-shot cross-lingual accuracy on XNLI and a 3.6%
improvement over XLM-R Base, creating a new SOTA on this benchmark. We have
made our pre-trained models and inference code publicly available at
https://github.com/microsoft/DeBERTa.
| no_new_dataset | 0.952926 |
2112.05282 | Quoc Viet Vo | Viet Quoc Vo and Ehsan Abbasnejad and Damith C. Ranasinghe | RamBoAttack: A Robust Query Efficient Deep Neural Network Decision
Exploit | Published in Network and Distributed System Security (NDSS) Symposium
2022. Code is available at https://ramboattack.github.io/ | null | 10.14722/ndss.2022.24200 | null | cs.LG cs.AI cs.CR cs.CV | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Machine learning models are critically susceptible to evasion attacks from
adversarial examples. Generally, adversarial examples, modified inputs
deceptively similar to the original input, are constructed under whitebox
settings by adversaries with full access to the model. However, recent attacks
have shown a remarkable reduction in query numbers to craft adversarial
examples using blackbox attacks. Particularly, alarming is the ability to
exploit the classification decision from the access interface of a trained
model provided by a growing number of Machine Learning as a Service providers
including Google, Microsoft, IBM and used by a plethora of applications
incorporating these models. The ability of an adversary to exploit only the
predicted label from a model to craft adversarial examples is distinguished as
a decision-based attack. In our study, we first deep dive into recent
state-of-the-art decision-based attacks in ICLR and SP to highlight the costly
nature of discovering low distortion adversarial employing gradient estimation
methods. We develop a robust query efficient attack capable of avoiding
entrapment in a local minimum and misdirection from noisy gradients seen in
gradient estimation methods. The attack method we propose, RamBoAttack,
exploits the notion of Randomized Block Coordinate Descent to explore the
hidden classifier manifold, targeting perturbations to manipulate only
localized input features to address the issues of gradient estimation methods.
Importantly, the RamBoAttack is more robust to the different sample inputs
available to an adversary and the targeted class. Overall, for a given target
class, RamBoAttack is demonstrated to be more robust at achieving a lower
distortion within a given query budget. We curate our extensive results using
the large-scale high-resolution ImageNet dataset and open-source our attack,
test samples and artifacts on GitHub.
| [
{
"version": "v1",
"created": "Fri, 10 Dec 2021 01:25:24 GMT"
},
{
"version": "v2",
"created": "Fri, 2 Dec 2022 01:30:02 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 02:16:07 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Vo",
"Viet Quoc",
""
],
[
"Abbasnejad",
"Ehsan",
""
],
[
"Ranasinghe",
"Damith C.",
""
]
] | TITLE: RamBoAttack: A Robust Query Efficient Deep Neural Network Decision
Exploit
ABSTRACT: Machine learning models are critically susceptible to evasion attacks from
adversarial examples. Generally, adversarial examples, modified inputs
deceptively similar to the original input, are constructed under whitebox
settings by adversaries with full access to the model. However, recent attacks
have shown a remarkable reduction in query numbers to craft adversarial
examples using blackbox attacks. Particularly, alarming is the ability to
exploit the classification decision from the access interface of a trained
model provided by a growing number of Machine Learning as a Service providers
including Google, Microsoft, IBM and used by a plethora of applications
incorporating these models. The ability of an adversary to exploit only the
predicted label from a model to craft adversarial examples is distinguished as
a decision-based attack. In our study, we first deep dive into recent
state-of-the-art decision-based attacks in ICLR and SP to highlight the costly
nature of discovering low distortion adversarial employing gradient estimation
methods. We develop a robust query efficient attack capable of avoiding
entrapment in a local minimum and misdirection from noisy gradients seen in
gradient estimation methods. The attack method we propose, RamBoAttack,
exploits the notion of Randomized Block Coordinate Descent to explore the
hidden classifier manifold, targeting perturbations to manipulate only
localized input features to address the issues of gradient estimation methods.
Importantly, the RamBoAttack is more robust to the different sample inputs
available to an adversary and the targeted class. Overall, for a given target
class, RamBoAttack is demonstrated to be more robust at achieving a lower
distortion within a given query budget. We curate our extensive results using
the large-scale high-resolution ImageNet dataset and open-source our attack,
test samples and artifacts on GitHub.
| no_new_dataset | 0.939025 |
2203.09382 | Claudio Gallicchio | Claudio Gallicchio | Euler State Networks: Non-dissipative Reservoir Computing | paper submitted to journal | null | null | null | cs.LG cs.AI stat.ML | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Inspired by the numerical solution of ordinary differential equations, in
this paper we propose a novel Reservoir Computing (RC) model, called the Euler
State Network (EuSN). The presented approach makes use of forward Euler
discretization and antisymmetric recurrent matrices to design reservoir
dynamics that are both stable and non-dissipative by construction.
Our mathematical analysis shows that the resulting model is biased towards a
unitary effective spectral radius and zero local Lyapunov exponents,
intrinsically operating near to the edge of stability. Experiments on long-term
memory tasks show the clear superiority of the proposed approach over standard
RC models in problems requiring effective propagation of input information over
multiple time-steps. Furthermore, results on time-series classification
benchmarks indicate that EuSN is able to match (or even exceed) the accuracy of
trainable Recurrent Neural Networks, while retaining the training efficiency of
the RC family, resulting in up to $\approx$ 490-fold savings in computation
time and $\approx$ 1750-fold savings in energy consumption.
| [
{
"version": "v1",
"created": "Thu, 17 Mar 2022 15:18:34 GMT"
},
{
"version": "v2",
"created": "Mon, 20 Feb 2023 16:49:34 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 15:18:21 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Gallicchio",
"Claudio",
""
]
] | TITLE: Euler State Networks: Non-dissipative Reservoir Computing
ABSTRACT: Inspired by the numerical solution of ordinary differential equations, in
this paper we propose a novel Reservoir Computing (RC) model, called the Euler
State Network (EuSN). The presented approach makes use of forward Euler
discretization and antisymmetric recurrent matrices to design reservoir
dynamics that are both stable and non-dissipative by construction.
Our mathematical analysis shows that the resulting model is biased towards a
unitary effective spectral radius and zero local Lyapunov exponents,
intrinsically operating near to the edge of stability. Experiments on long-term
memory tasks show the clear superiority of the proposed approach over standard
RC models in problems requiring effective propagation of input information over
multiple time-steps. Furthermore, results on time-series classification
benchmarks indicate that EuSN is able to match (or even exceed) the accuracy of
trainable Recurrent Neural Networks, while retaining the training efficiency of
the RC family, resulting in up to $\approx$ 490-fold savings in computation
time and $\approx$ 1750-fold savings in energy consumption.
| no_new_dataset | 0.949342 |
2203.11009 | Lukas Hedegaard | Lukas Hedegaard and Negar Heidari and Alexandros Iosifidis | Continual Spatio-Temporal Graph Convolutional Networks | 12 pages, 6 figures, 5 tables | Pattern Recognition 140 (2023) 109528 | 10.1016/j.patcog.2023.109528 | null | cs.CV cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Graph-based reasoning over skeleton data has emerged as a promising approach
for human action recognition. However, the application of prior graph-based
methods, which predominantly employ whole temporal sequences as their input, to
the setting of online inference entails considerable computational redundancy.
In this paper, we tackle this issue by reformulating the Spatio-Temporal Graph
Convolutional Neural Network as a Continual Inference Network, which can
perform step-by-step predictions in time without repeat frame processing. To
evaluate our method, we create a continual version of ST-GCN, CoST-GCN,
alongside two derived methods with different self-attention mechanisms, CoAGCN
and CoS-TR. We investigate weight transfer strategies and architectural
modifications for inference acceleration, and perform experiments on the NTU
RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400 datasets. Retaining similar
predictive accuracy, we observe up to 109x reduction in time complexity,
on-hardware accelerations of 26x, and reductions in maximum allocated memory of
52% during online inference.
| [
{
"version": "v1",
"created": "Mon, 21 Mar 2022 14:23:18 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 08:32:24 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Hedegaard",
"Lukas",
""
],
[
"Heidari",
"Negar",
""
],
[
"Iosifidis",
"Alexandros",
""
]
] | TITLE: Continual Spatio-Temporal Graph Convolutional Networks
ABSTRACT: Graph-based reasoning over skeleton data has emerged as a promising approach
for human action recognition. However, the application of prior graph-based
methods, which predominantly employ whole temporal sequences as their input, to
the setting of online inference entails considerable computational redundancy.
In this paper, we tackle this issue by reformulating the Spatio-Temporal Graph
Convolutional Neural Network as a Continual Inference Network, which can
perform step-by-step predictions in time without repeat frame processing. To
evaluate our method, we create a continual version of ST-GCN, CoST-GCN,
alongside two derived methods with different self-attention mechanisms, CoAGCN
and CoS-TR. We investigate weight transfer strategies and architectural
modifications for inference acceleration, and perform experiments on the NTU
RGB+D 60, NTU RGB+D 120, and Kinetics Skeleton 400 datasets. Retaining similar
predictive accuracy, we observe up to 109x reduction in time complexity,
on-hardware accelerations of 26x, and reductions in maximum allocated memory of
52% during online inference.
| no_new_dataset | 0.949201 |
2205.10636 | Xingzhe He | Xingzhe He, Bastian Wandt, Helge Rhodin | AutoLink: Self-supervised Learning of Human Skeletons and Object
Outlines by Linking Keypoints | NeurIPS 2022 | Advances in Neural Information Processing Systems 2022 | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Structured representations such as keypoints are widely used in pose
transfer, conditional image generation, animation, and 3D reconstruction.
However, their supervised learning requires expensive annotation for each
target domain. We propose a self-supervised method that learns to disentangle
object structure from the appearance with a graph of 2D keypoints linked by
straight edges. Both the keypoint location and their pairwise edge weights are
learned, given only a collection of images depicting the same object class. The
resulting graph is interpretable, for example, AutoLink recovers the human
skeleton topology when applied to images showing people. Our key ingredients
are i) an encoder that predicts keypoint locations in an input image, ii) a
shared graph as a latent variable that links the same pairs of keypoints in
every image, iii) an intermediate edge map that combines the latent graph edge
weights and keypoint locations in a soft, differentiable manner, and iv) an
inpainting objective on randomly masked images. Although simpler, AutoLink
outperforms existing self-supervised methods on the established keypoint and
pose estimation benchmarks and paves the way for structure-conditioned
generative models on more diverse datasets. Project website:
https://xingzhehe.github.io/autolink/.
| [
{
"version": "v1",
"created": "Sat, 21 May 2022 16:32:34 GMT"
},
{
"version": "v2",
"created": "Fri, 1 Jul 2022 08:28:05 GMT"
},
{
"version": "v3",
"created": "Fri, 16 Sep 2022 23:08:19 GMT"
},
{
"version": "v4",
"created": "Sat, 24 Sep 2022 01:50:36 GMT"
},
{
"version": "v5",
"created": "Sun, 9 Oct 2022 00:44:09 GMT"
},
{
"version": "v6",
"created": "Thu, 23 Mar 2023 18:31:48 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"He",
"Xingzhe",
""
],
[
"Wandt",
"Bastian",
""
],
[
"Rhodin",
"Helge",
""
]
] | TITLE: AutoLink: Self-supervised Learning of Human Skeletons and Object
Outlines by Linking Keypoints
ABSTRACT: Structured representations such as keypoints are widely used in pose
transfer, conditional image generation, animation, and 3D reconstruction.
However, their supervised learning requires expensive annotation for each
target domain. We propose a self-supervised method that learns to disentangle
object structure from the appearance with a graph of 2D keypoints linked by
straight edges. Both the keypoint location and their pairwise edge weights are
learned, given only a collection of images depicting the same object class. The
resulting graph is interpretable, for example, AutoLink recovers the human
skeleton topology when applied to images showing people. Our key ingredients
are i) an encoder that predicts keypoint locations in an input image, ii) a
shared graph as a latent variable that links the same pairs of keypoints in
every image, iii) an intermediate edge map that combines the latent graph edge
weights and keypoint locations in a soft, differentiable manner, and iv) an
inpainting objective on randomly masked images. Although simpler, AutoLink
outperforms existing self-supervised methods on the established keypoint and
pose estimation benchmarks and paves the way for structure-conditioned
generative models on more diverse datasets. Project website:
https://xingzhehe.github.io/autolink/.
| no_new_dataset | 0.948822 |
2206.07018 | Si Chen | Si Chen, Yi Zeng, Jiachen T.Wang, Won Park, Xun Chen, Lingjuan Lyu,
Zhuoqing Mao, Ruoxi Jia | Turning a Curse into a Blessing: Enabling In-Distribution-Data-Free
Backdoor Removal via Stabilized Model Inversion | Because of an equation and author informational error, this paper has
been withdrawn by the submitter | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Many backdoor removal techniques in machine learning models require clean
in-distribution data, which may not always be available due to proprietary
datasets. Model inversion techniques, often considered privacy threats, can
reconstruct realistic training samples, potentially eliminating the need for
in-distribution data. Prior attempts to combine backdoor removal and model
inversion yielded limited results. Our work is the first to provide a thorough
understanding of leveraging model inversion for effective backdoor removal by
addressing key questions about reconstructed samples' properties, perceptual
similarity, and the potential presence of backdoor triggers.
We establish that relying solely on perceptual similarity is insufficient for
robust defenses, and the stability of model predictions in response to input
and parameter perturbations is also crucial. To tackle this, we introduce a
novel bi-level optimization-based framework for model inversion, promoting
stability and visual quality. Interestingly, we discover that reconstructed
samples from a pre-trained generator's latent space are backdoor-free, even
when utilizing signals from a backdoored model. We provide a theoretical
analysis to support this finding. Our evaluation demonstrates that our
stabilized model inversion technique achieves state-of-the-art backdoor removal
performance without clean in-distribution data, matching or surpassing
performance using the same amount of clean samples.
| [
{
"version": "v1",
"created": "Tue, 14 Jun 2022 17:32:04 GMT"
},
{
"version": "v2",
"created": "Fri, 15 Jul 2022 18:14:07 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 01:32:49 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Chen",
"Si",
""
],
[
"Zeng",
"Yi",
""
],
[
"Wang",
"Jiachen T.",
""
],
[
"Park",
"Won",
""
],
[
"Chen",
"Xun",
""
],
[
"Lyu",
"Lingjuan",
""
],
[
"Mao",
"Zhuoqing",
""
],
[
"Jia",
"Ruoxi",
""
]
] | TITLE: Turning a Curse into a Blessing: Enabling In-Distribution-Data-Free
Backdoor Removal via Stabilized Model Inversion
ABSTRACT: Many backdoor removal techniques in machine learning models require clean
in-distribution data, which may not always be available due to proprietary
datasets. Model inversion techniques, often considered privacy threats, can
reconstruct realistic training samples, potentially eliminating the need for
in-distribution data. Prior attempts to combine backdoor removal and model
inversion yielded limited results. Our work is the first to provide a thorough
understanding of leveraging model inversion for effective backdoor removal by
addressing key questions about reconstructed samples' properties, perceptual
similarity, and the potential presence of backdoor triggers.
We establish that relying solely on perceptual similarity is insufficient for
robust defenses, and the stability of model predictions in response to input
and parameter perturbations is also crucial. To tackle this, we introduce a
novel bi-level optimization-based framework for model inversion, promoting
stability and visual quality. Interestingly, we discover that reconstructed
samples from a pre-trained generator's latent space are backdoor-free, even
when utilizing signals from a backdoored model. We provide a theoretical
analysis to support this finding. Our evaluation demonstrates that our
stabilized model inversion technique achieves state-of-the-art backdoor removal
performance without clean in-distribution data, matching or surpassing
performance using the same amount of clean samples.
| no_new_dataset | 0.945801 |
2206.11896 | Viktor Rudnev | Viktor Rudnev and Mohamed Elgharib and Christian Theobalt and
Vladislav Golyanik | EventNeRF: Neural Radiance Fields from a Single Colour Event Camera | 19 pages, 21 figures, 3 tables; CVPR 2023 | Computer Vision and Pattern Recognition (CVPR) 2023 | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Asynchronously operating event cameras find many applications due to their
high dynamic range, vanishingly low motion blur, low latency and low data
bandwidth. The field saw remarkable progress during the last few years, and
existing event-based 3D reconstruction approaches recover sparse point clouds
of the scene. However, such sparsity is a limiting factor in many cases,
especially in computer vision and graphics, that has not been addressed
satisfactorily so far. Accordingly, this paper proposes the first approach for
3D-consistent, dense and photorealistic novel view synthesis using just a
single colour event stream as input. At its core is a neural radiance field
trained entirely in a self-supervised manner from events while preserving the
original resolution of the colour event channels. Next, our ray sampling
strategy is tailored to events and allows for data-efficient training. At test,
our method produces results in the RGB space at unprecedented quality. We
evaluate our method qualitatively and numerically on several challenging
synthetic and real scenes and show that it produces significantly denser and
more visually appealing renderings than the existing methods. We also
demonstrate robustness in challenging scenarios with fast motion and under low
lighting conditions. We release the newly recorded dataset and our source code
to facilitate the research field, see https://4dqv.mpi-inf.mpg.de/EventNeRF.
| [
{
"version": "v1",
"created": "Thu, 23 Jun 2022 17:59:53 GMT"
},
{
"version": "v2",
"created": "Mon, 28 Nov 2022 17:00:06 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 16:57:36 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Rudnev",
"Viktor",
""
],
[
"Elgharib",
"Mohamed",
""
],
[
"Theobalt",
"Christian",
""
],
[
"Golyanik",
"Vladislav",
""
]
] | TITLE: EventNeRF: Neural Radiance Fields from a Single Colour Event Camera
ABSTRACT: Asynchronously operating event cameras find many applications due to their
high dynamic range, vanishingly low motion blur, low latency and low data
bandwidth. The field saw remarkable progress during the last few years, and
existing event-based 3D reconstruction approaches recover sparse point clouds
of the scene. However, such sparsity is a limiting factor in many cases,
especially in computer vision and graphics, that has not been addressed
satisfactorily so far. Accordingly, this paper proposes the first approach for
3D-consistent, dense and photorealistic novel view synthesis using just a
single colour event stream as input. At its core is a neural radiance field
trained entirely in a self-supervised manner from events while preserving the
original resolution of the colour event channels. Next, our ray sampling
strategy is tailored to events and allows for data-efficient training. At test,
our method produces results in the RGB space at unprecedented quality. We
evaluate our method qualitatively and numerically on several challenging
synthetic and real scenes and show that it produces significantly denser and
more visually appealing renderings than the existing methods. We also
demonstrate robustness in challenging scenarios with fast motion and under low
lighting conditions. We release the newly recorded dataset and our source code
to facilitate the research field, see https://4dqv.mpi-inf.mpg.de/EventNeRF.
| new_dataset | 0.954774 |
2207.10660 | Garrick Brazil | Garrick Brazil, Abhinav Kumar, Julian Straub, Nikhila Ravi, Justin
Johnson, Georgia Gkioxari | Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild | CVPR 2023, Project website: https://omni3d.garrickbrazil.com/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Recognizing scenes and objects in 3D from a single image is a longstanding
goal of computer vision with applications in robotics and AR/VR. For 2D
recognition, large datasets and scalable solutions have led to unprecedented
advances. In 3D, existing benchmarks are small in size and approaches
specialize in few object categories and specific domains, e.g. urban driving
scenes. Motivated by the success of 2D recognition, we revisit the task of 3D
object detection by introducing a large benchmark, called Omni3D. Omni3D
re-purposes and combines existing datasets resulting in 234k images annotated
with more than 3 million instances and 98 categories. 3D detection at such
scale is challenging due to variations in camera intrinsics and the rich
diversity of scene and object types. We propose a model, called Cube R-CNN,
designed to generalize across camera and scene types with a unified approach.
We show that Cube R-CNN outperforms prior works on the larger Omni3D and
existing benchmarks. Finally, we prove that Omni3D is a powerful dataset for 3D
object recognition and show that it improves single-dataset performance and can
accelerate learning on new smaller datasets via pre-training.
| [
{
"version": "v1",
"created": "Thu, 21 Jul 2022 17:56:22 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 00:42:18 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Brazil",
"Garrick",
""
],
[
"Kumar",
"Abhinav",
""
],
[
"Straub",
"Julian",
""
],
[
"Ravi",
"Nikhila",
""
],
[
"Johnson",
"Justin",
""
],
[
"Gkioxari",
"Georgia",
""
]
] | TITLE: Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild
ABSTRACT: Recognizing scenes and objects in 3D from a single image is a longstanding
goal of computer vision with applications in robotics and AR/VR. For 2D
recognition, large datasets and scalable solutions have led to unprecedented
advances. In 3D, existing benchmarks are small in size and approaches
specialize in few object categories and specific domains, e.g. urban driving
scenes. Motivated by the success of 2D recognition, we revisit the task of 3D
object detection by introducing a large benchmark, called Omni3D. Omni3D
re-purposes and combines existing datasets resulting in 234k images annotated
with more than 3 million instances and 98 categories. 3D detection at such
scale is challenging due to variations in camera intrinsics and the rich
diversity of scene and object types. We propose a model, called Cube R-CNN,
designed to generalize across camera and scene types with a unified approach.
We show that Cube R-CNN outperforms prior works on the larger Omni3D and
existing benchmarks. Finally, we prove that Omni3D is a powerful dataset for 3D
object recognition and show that it improves single-dataset performance and can
accelerate learning on new smaller datasets via pre-training.
| no_new_dataset | 0.83868 |
2207.14251 | Yanai Elazar | Yanai Elazar, Nora Kassner, Shauli Ravfogel, Amir Feder, Abhilasha
Ravichander, Marius Mosbach, Yonatan Belinkov, Hinrich Sch\"utze, Yoav
Goldberg | Measuring Causal Effects of Data Statistics on Language Model's
`Factual' Predictions | We received a criticism regarding the validity of the causal
formulation in this paper. We will address them in an upcoming version | null | null | null | cs.CL | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large amounts of training data are one of the major reasons for the high
performance of state-of-the-art NLP models. But what exactly in the training
data causes a model to make a certain prediction? We seek to answer this
question by providing a language for describing how training data influences
predictions, through a causal framework. Importantly, our framework bypasses
the need to retrain expensive models and allows us to estimate causal effects
based on observational data alone. Addressing the problem of extracting factual
knowledge from pretrained language models (PLMs), we focus on simple data
statistics such as co-occurrence counts and show that these statistics do
influence the predictions of PLMs, suggesting that such models rely on shallow
heuristics. Our causal framework and our results demonstrate the importance of
studying datasets and the benefits of causality for understanding NLP models.
| [
{
"version": "v1",
"created": "Thu, 28 Jul 2022 17:36:24 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 07:18:59 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Elazar",
"Yanai",
""
],
[
"Kassner",
"Nora",
""
],
[
"Ravfogel",
"Shauli",
""
],
[
"Feder",
"Amir",
""
],
[
"Ravichander",
"Abhilasha",
""
],
[
"Mosbach",
"Marius",
""
],
[
"Belinkov",
"Yonatan",
""
],
[
"Schütze",
"Hinrich",
""
],
[
"Goldberg",
"Yoav",
""
]
] | TITLE: Measuring Causal Effects of Data Statistics on Language Model's
`Factual' Predictions
ABSTRACT: Large amounts of training data are one of the major reasons for the high
performance of state-of-the-art NLP models. But what exactly in the training
data causes a model to make a certain prediction? We seek to answer this
question by providing a language for describing how training data influences
predictions, through a causal framework. Importantly, our framework bypasses
the need to retrain expensive models and allows us to estimate causal effects
based on observational data alone. Addressing the problem of extracting factual
knowledge from pretrained language models (PLMs), we focus on simple data
statistics such as co-occurrence counts and show that these statistics do
influence the predictions of PLMs, suggesting that such models rely on shallow
heuristics. Our causal framework and our results demonstrate the importance of
studying datasets and the benefits of causality for understanding NLP models.
| no_new_dataset | 0.952353 |
2210.01936 | Federico Bianchi | Mert Yuksekgonul, Federico Bianchi, Pratyusha Kalluri, Dan Jurafsky,
James Zou | When and why vision-language models behave like bags-of-words, and what
to do about it? | ICLR 2023 Oral (notable-top-5%) | null | null | null | cs.CV cs.AI cs.CL cs.LG | http://creativecommons.org/licenses/by/4.0/ | Despite the success of large vision and language models (VLMs) in many
downstream applications, it is unclear how well they encode compositional
information. Here, we create the Attribution, Relation, and Order (ARO)
benchmark to systematically evaluate the ability of VLMs to understand
different types of relationships, attributes, and order. ARO consists of Visual
Genome Attribution, to test the understanding of objects' properties; Visual
Genome Relation, to test for relational understanding; and COCO &
Flickr30k-Order, to test for order sensitivity. ARO is orders of magnitude
larger than previous benchmarks of compositionality, with more than 50,000 test
cases. We show where state-of-the-art VLMs have poor relational understanding,
can blunder when linking objects to their attributes, and demonstrate a severe
lack of order sensitivity. VLMs are predominantly trained and evaluated on
large datasets with rich compositional structure in the images and captions.
Yet, training on these datasets has not been enough to address the lack of
compositional understanding, and evaluating on these datasets has failed to
surface this deficiency. To understand why these limitations emerge and are not
represented in the standard tests, we zoom into the evaluation and training
procedures. We demonstrate that it is possible to perform well on retrieval
over existing datasets without using the composition and order information.
Given that contrastive pretraining optimizes for retrieval on datasets with
similar shortcuts, we hypothesize that this can explain why the models do not
need to learn to represent compositional information. This finding suggests a
natural solution: composition-aware hard negative mining. We show that a
simple-to-implement modification of contrastive learning significantly improves
the performance on tasks requiring understanding of order and compositionality.
| [
{
"version": "v1",
"created": "Tue, 4 Oct 2022 22:13:25 GMT"
},
{
"version": "v2",
"created": "Thu, 6 Oct 2022 01:11:39 GMT"
},
{
"version": "v3",
"created": "Thu, 23 Mar 2023 23:21:38 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Yuksekgonul",
"Mert",
""
],
[
"Bianchi",
"Federico",
""
],
[
"Kalluri",
"Pratyusha",
""
],
[
"Jurafsky",
"Dan",
""
],
[
"Zou",
"James",
""
]
] | TITLE: When and why vision-language models behave like bags-of-words, and what
to do about it?
ABSTRACT: Despite the success of large vision and language models (VLMs) in many
downstream applications, it is unclear how well they encode compositional
information. Here, we create the Attribution, Relation, and Order (ARO)
benchmark to systematically evaluate the ability of VLMs to understand
different types of relationships, attributes, and order. ARO consists of Visual
Genome Attribution, to test the understanding of objects' properties; Visual
Genome Relation, to test for relational understanding; and COCO &
Flickr30k-Order, to test for order sensitivity. ARO is orders of magnitude
larger than previous benchmarks of compositionality, with more than 50,000 test
cases. We show where state-of-the-art VLMs have poor relational understanding,
can blunder when linking objects to their attributes, and demonstrate a severe
lack of order sensitivity. VLMs are predominantly trained and evaluated on
large datasets with rich compositional structure in the images and captions.
Yet, training on these datasets has not been enough to address the lack of
compositional understanding, and evaluating on these datasets has failed to
surface this deficiency. To understand why these limitations emerge and are not
represented in the standard tests, we zoom into the evaluation and training
procedures. We demonstrate that it is possible to perform well on retrieval
over existing datasets without using the composition and order information.
Given that contrastive pretraining optimizes for retrieval on datasets with
similar shortcuts, we hypothesize that this can explain why the models do not
need to learn to represent compositional information. This finding suggests a
natural solution: composition-aware hard negative mining. We show that a
simple-to-implement modification of contrastive learning significantly improves
the performance on tasks requiring understanding of order and compositionality.
| no_new_dataset | 0.950641 |
2210.12685 | Min Yang | Yufeng Wang, Cong Xu, Min Yang, Jin Zhang | Less Emphasis on Difficult Layer Regions: Curriculum Learning for
Singularly Perturbed Convection-Diffusion-Reaction Problems | 22 pages | null | null | null | cs.LG cs.NA math.NA | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Although Physics-Informed Neural Networks (PINNs) have been successfully
applied in a wide variety of science and engineering fields, they can fail to
accurately predict the underlying solution in slightly challenging
convection-diffusion-reaction problems. In this paper, we investigate the
reason of this failure from a domain distribution perspective, and identify
that learning multi-scale fields simultaneously makes the network unable to
advance its training and easily get stuck in poor local minima. We show that
the widespread experience of sampling more collocation points in high-loss
layer regions hardly help optimize and may even worsen the results. These
findings motivate the development of a novel curriculum learning method that
encourages neural networks to prioritize learning on easier non-layer regions
while downplaying learning on harder layer regions. The proposed method helps
PINNs automatically adjust the learning emphasis and thereby facilitate the
optimization procedure. Numerical results on typical benchmark equations show
that the proposed curriculum learning approach mitigates the failure modes of
PINNs and can produce accurate results for very sharp boundary and interior
layers. Our work reveals that for equations whose solutions have large scale
differences, paying less attention to high-loss regions can be an effective
strategy for learning them accurately.
| [
{
"version": "v1",
"created": "Sun, 23 Oct 2022 10:15:32 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 10:09:01 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Wang",
"Yufeng",
""
],
[
"Xu",
"Cong",
""
],
[
"Yang",
"Min",
""
],
[
"Zhang",
"Jin",
""
]
] | TITLE: Less Emphasis on Difficult Layer Regions: Curriculum Learning for
Singularly Perturbed Convection-Diffusion-Reaction Problems
ABSTRACT: Although Physics-Informed Neural Networks (PINNs) have been successfully
applied in a wide variety of science and engineering fields, they can fail to
accurately predict the underlying solution in slightly challenging
convection-diffusion-reaction problems. In this paper, we investigate the
reason of this failure from a domain distribution perspective, and identify
that learning multi-scale fields simultaneously makes the network unable to
advance its training and easily get stuck in poor local minima. We show that
the widespread experience of sampling more collocation points in high-loss
layer regions hardly help optimize and may even worsen the results. These
findings motivate the development of a novel curriculum learning method that
encourages neural networks to prioritize learning on easier non-layer regions
while downplaying learning on harder layer regions. The proposed method helps
PINNs automatically adjust the learning emphasis and thereby facilitate the
optimization procedure. Numerical results on typical benchmark equations show
that the proposed curriculum learning approach mitigates the failure modes of
PINNs and can produce accurate results for very sharp boundary and interior
layers. Our work reveals that for equations whose solutions have large scale
differences, paying less attention to high-loss regions can be an effective
strategy for learning them accurately.
| no_new_dataset | 0.948106 |
2211.08834 | Miran Heo | Miran Heo, Sukjun Hwang, Jeongseok Hyun, Hanjung Kim, Seoung Wug Oh,
Joon-Young Lee, Seon Joo Kim | A Generalized Framework for Video Instance Segmentation | CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | The handling of long videos with complex and occluded sequences has recently
emerged as a new challenge in the video instance segmentation (VIS) community.
However, existing methods have limitations in addressing this challenge. We
argue that the biggest bottleneck in current approaches is the discrepancy
between training and inference. To effectively bridge this gap, we propose a
Generalized framework for VIS, namely GenVIS, that achieves state-of-the-art
performance on challenging benchmarks without designing complicated
architectures or requiring extra post-processing. The key contribution of
GenVIS is the learning strategy, which includes a query-based training pipeline
for sequential learning with a novel target label assignment. Additionally, we
introduce a memory that effectively acquires information from previous states.
Thanks to the new perspective, which focuses on building relationships between
separate frames or clips, GenVIS can be flexibly executed in both online and
semi-online manner. We evaluate our approach on popular VIS benchmarks,
achieving state-of-the-art results on YouTube-VIS 2019/2021/2022 and Occluded
VIS (OVIS). Notably, we greatly outperform the state-of-the-art on the long VIS
benchmark (OVIS), improving 5.6 AP with ResNet-50 backbone. Code is available
at https://github.com/miranheo/GenVIS.
| [
{
"version": "v1",
"created": "Wed, 16 Nov 2022 11:17:19 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 15:26:13 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Heo",
"Miran",
""
],
[
"Hwang",
"Sukjun",
""
],
[
"Hyun",
"Jeongseok",
""
],
[
"Kim",
"Hanjung",
""
],
[
"Oh",
"Seoung Wug",
""
],
[
"Lee",
"Joon-Young",
""
],
[
"Kim",
"Seon Joo",
""
]
] | TITLE: A Generalized Framework for Video Instance Segmentation
ABSTRACT: The handling of long videos with complex and occluded sequences has recently
emerged as a new challenge in the video instance segmentation (VIS) community.
However, existing methods have limitations in addressing this challenge. We
argue that the biggest bottleneck in current approaches is the discrepancy
between training and inference. To effectively bridge this gap, we propose a
Generalized framework for VIS, namely GenVIS, that achieves state-of-the-art
performance on challenging benchmarks without designing complicated
architectures or requiring extra post-processing. The key contribution of
GenVIS is the learning strategy, which includes a query-based training pipeline
for sequential learning with a novel target label assignment. Additionally, we
introduce a memory that effectively acquires information from previous states.
Thanks to the new perspective, which focuses on building relationships between
separate frames or clips, GenVIS can be flexibly executed in both online and
semi-online manner. We evaluate our approach on popular VIS benchmarks,
achieving state-of-the-art results on YouTube-VIS 2019/2021/2022 and Occluded
VIS (OVIS). Notably, we greatly outperform the state-of-the-art on the long VIS
benchmark (OVIS), improving 5.6 AP with ResNet-50 backbone. Code is available
at https://github.com/miranheo/GenVIS.
| no_new_dataset | 0.9463 |
2211.10277 | Tao Yu | Tao Yu, Zhihe Lu, Xin Jin, Zhibo Chen, Xinchao Wang | Task Residual for Tuning Vision-Language Models | Accepted to CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Large-scale vision-language models (VLMs) pre-trained on billion-level data
have learned general visual representations and broad visual concepts. In
principle, the well-learned knowledge structure of the VLMs should be inherited
appropriately when being transferred to downstream tasks with limited data.
However, most existing efficient transfer learning (ETL) approaches for VLMs
either damage or are excessively biased towards the prior knowledge, e.g.,
prompt tuning (PT) discards the pre-trained text-based classifier and builds a
new one while adapter-style tuning (AT) fully relies on the pre-trained
features. To address this, we propose a new efficient tuning approach for VLMs
named Task Residual Tuning (TaskRes), which performs directly on the text-based
classifier and explicitly decouples the prior knowledge of the pre-trained
models and new knowledge regarding a target task. Specifically, TaskRes keeps
the original classifier weights from the VLMs frozen and obtains a new
classifier for the target task by tuning a set of prior-independent parameters
as a residual to the original one, which enables reliable prior knowledge
preservation and flexible task-specific knowledge exploration. The proposed
TaskRes is simple yet effective, which significantly outperforms previous ETL
methods (e.g., PT and AT) on 11 benchmark datasets while requiring minimal
effort for the implementation. Our code is available at
https://github.com/geekyutao/TaskRes.
| [
{
"version": "v1",
"created": "Fri, 18 Nov 2022 15:09:03 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 17:22:32 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Yu",
"Tao",
""
],
[
"Lu",
"Zhihe",
""
],
[
"Jin",
"Xin",
""
],
[
"Chen",
"Zhibo",
""
],
[
"Wang",
"Xinchao",
""
]
] | TITLE: Task Residual for Tuning Vision-Language Models
ABSTRACT: Large-scale vision-language models (VLMs) pre-trained on billion-level data
have learned general visual representations and broad visual concepts. In
principle, the well-learned knowledge structure of the VLMs should be inherited
appropriately when being transferred to downstream tasks with limited data.
However, most existing efficient transfer learning (ETL) approaches for VLMs
either damage or are excessively biased towards the prior knowledge, e.g.,
prompt tuning (PT) discards the pre-trained text-based classifier and builds a
new one while adapter-style tuning (AT) fully relies on the pre-trained
features. To address this, we propose a new efficient tuning approach for VLMs
named Task Residual Tuning (TaskRes), which performs directly on the text-based
classifier and explicitly decouples the prior knowledge of the pre-trained
models and new knowledge regarding a target task. Specifically, TaskRes keeps
the original classifier weights from the VLMs frozen and obtains a new
classifier for the target task by tuning a set of prior-independent parameters
as a residual to the original one, which enables reliable prior knowledge
preservation and flexible task-specific knowledge exploration. The proposed
TaskRes is simple yet effective, which significantly outperforms previous ETL
methods (e.g., PT and AT) on 11 benchmark datasets while requiring minimal
effort for the implementation. Our code is available at
https://github.com/geekyutao/TaskRes.
| no_new_dataset | 0.950549 |
2211.12380 | Eloi Zablocki | Mehdi Zemni, Micka\"el Chen, \'Eloi Zablocki, H\'edi Ben-Younes,
Patrick P\'erez, Matthieu Cord | OCTET: Object-aware Counterfactual Explanations | CVPR 2023 | null | null | null | cs.CV cs.AI | http://creativecommons.org/licenses/by-nc-nd/4.0/ | Nowadays, deep vision models are being widely deployed in safety-critical
applications, e.g., autonomous driving, and explainability of such models is
becoming a pressing concern. Among explanation methods, counterfactual
explanations aim to find minimal and interpretable changes to the input image
that would also change the output of the model to be explained. Such
explanations point end-users at the main factors that impact the decision of
the model. However, previous methods struggle to explain decision models
trained on images with many objects, e.g., urban scenes, which are more
difficult to work with but also arguably more critical to explain. In this
work, we propose to tackle this issue with an object-centric framework for
counterfactual explanation generation. Our method, inspired by recent
generative modeling works, encodes the query image into a latent space that is
structured in a way to ease object-level manipulations. Doing so, it provides
the end-user with control over which search directions (e.g., spatial
displacement of objects, style modification, etc.) are to be explored during
the counterfactual generation. We conduct a set of experiments on
counterfactual explanation benchmarks for driving scenes, and we show that our
method can be adapted beyond classification, e.g., to explain semantic
segmentation models. To complete our analysis, we design and run a user study
that measures the usefulness of counterfactual explanations in understanding a
decision model. Code is available at https://github.com/valeoai/OCTET.
| [
{
"version": "v1",
"created": "Tue, 22 Nov 2022 16:23:12 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 16:01:24 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Zemni",
"Mehdi",
""
],
[
"Chen",
"Mickaël",
""
],
[
"Zablocki",
"Éloi",
""
],
[
"Ben-Younes",
"Hédi",
""
],
[
"Pérez",
"Patrick",
""
],
[
"Cord",
"Matthieu",
""
]
] | TITLE: OCTET: Object-aware Counterfactual Explanations
ABSTRACT: Nowadays, deep vision models are being widely deployed in safety-critical
applications, e.g., autonomous driving, and explainability of such models is
becoming a pressing concern. Among explanation methods, counterfactual
explanations aim to find minimal and interpretable changes to the input image
that would also change the output of the model to be explained. Such
explanations point end-users at the main factors that impact the decision of
the model. However, previous methods struggle to explain decision models
trained on images with many objects, e.g., urban scenes, which are more
difficult to work with but also arguably more critical to explain. In this
work, we propose to tackle this issue with an object-centric framework for
counterfactual explanation generation. Our method, inspired by recent
generative modeling works, encodes the query image into a latent space that is
structured in a way to ease object-level manipulations. Doing so, it provides
the end-user with control over which search directions (e.g., spatial
displacement of objects, style modification, etc.) are to be explored during
the counterfactual generation. We conduct a set of experiments on
counterfactual explanation benchmarks for driving scenes, and we show that our
method can be adapted beyond classification, e.g., to explain semantic
segmentation models. To complete our analysis, we design and run a user study
that measures the usefulness of counterfactual explanations in understanding a
decision model. Code is available at https://github.com/valeoai/OCTET.
| no_new_dataset | 0.943556 |
2211.13874 | Haoran Bai | Haoran Bai, Di Kang, Haoxian Zhang, Jinshan Pan, Linchao Bao | FFHQ-UV: Normalized Facial UV-Texture Dataset for 3D Face Reconstruction | The dataset, code, and pre-trained texture decoder are publicly
available at https://github.com/csbhr/FFHQ-UV | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | We present a large-scale facial UV-texture dataset that contains over 50,000
high-quality texture UV-maps with even illuminations, neutral expressions, and
cleaned facial regions, which are desired characteristics for rendering
realistic 3D face models under different lighting conditions. The dataset is
derived from a large-scale face image dataset namely FFHQ, with the help of our
fully automatic and robust UV-texture production pipeline. Our pipeline
utilizes the recent advances in StyleGAN-based facial image editing approaches
to generate multi-view normalized face images from single-image inputs. An
elaborated UV-texture extraction, correction, and completion procedure is then
applied to produce high-quality UV-maps from the normalized face images.
Compared with existing UV-texture datasets, our dataset has more diverse and
higher-quality texture maps. We further train a GAN-based texture decoder as
the nonlinear texture basis for parametric fitting based 3D face
reconstruction. Experiments show that our method improves the reconstruction
accuracy over state-of-the-art approaches, and more importantly, produces
high-quality texture maps that are ready for realistic renderings. The dataset,
code, and pre-trained texture decoder are publicly available at
https://github.com/csbhr/FFHQ-UV.
| [
{
"version": "v1",
"created": "Fri, 25 Nov 2022 03:21:05 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 14:44:50 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Bai",
"Haoran",
""
],
[
"Kang",
"Di",
""
],
[
"Zhang",
"Haoxian",
""
],
[
"Pan",
"Jinshan",
""
],
[
"Bao",
"Linchao",
""
]
] | TITLE: FFHQ-UV: Normalized Facial UV-Texture Dataset for 3D Face Reconstruction
ABSTRACT: We present a large-scale facial UV-texture dataset that contains over 50,000
high-quality texture UV-maps with even illuminations, neutral expressions, and
cleaned facial regions, which are desired characteristics for rendering
realistic 3D face models under different lighting conditions. The dataset is
derived from a large-scale face image dataset namely FFHQ, with the help of our
fully automatic and robust UV-texture production pipeline. Our pipeline
utilizes the recent advances in StyleGAN-based facial image editing approaches
to generate multi-view normalized face images from single-image inputs. An
elaborated UV-texture extraction, correction, and completion procedure is then
applied to produce high-quality UV-maps from the normalized face images.
Compared with existing UV-texture datasets, our dataset has more diverse and
higher-quality texture maps. We further train a GAN-based texture decoder as
the nonlinear texture basis for parametric fitting based 3D face
reconstruction. Experiments show that our method improves the reconstruction
accuracy over state-of-the-art approaches, and more importantly, produces
high-quality texture maps that are ready for realistic renderings. The dataset,
code, and pre-trained texture decoder are publicly available at
https://github.com/csbhr/FFHQ-UV.
| new_dataset | 0.971483 |
2211.15951 | HyeonCheol Moon | HyeonCheol Moon, JinWoo Jeong, SungJei Kim | Feature-domain Adaptive Contrastive Distillation for Efficient Single
Image Super-Resolution | Under review | null | null | null | cs.CV cs.CE | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Recently, CNN-based SISR has numerous parameters and high computational cost
to achieve better performance, limiting its applicability to
resource-constrained devices such as mobile. As one of the methods to make the
network efficient, Knowledge Distillation (KD), which transfers teacher's
useful knowledge to student, is currently being studied. More recently, KD for
SISR utilizes Feature Distillation (FD) to minimize the Euclidean distance loss
of feature maps between teacher and student networks, but it does not
sufficiently consider how to effectively and meaningfully deliver knowledge
from teacher to improve the student performance at given network capacity
constraints. In this paper, we propose a feature-domain adaptive contrastive
distillation (FACD) method for efficiently training lightweight student SISR
networks. We show the limitations of the existing FD methods using Euclidean
distance loss, and propose a feature-domain contrastive loss that makes a
student network learn richer information from the teacher's representation in
the feature domain. In addition, we propose an adaptive distillation that
selectively applies distillation depending on the conditions of the training
patches. The experimental results show that the student EDSR and RCAN networks
with the proposed FACD scheme improves not only the PSNR performance of the
entire benchmark datasets and scales, but also the subjective image quality
compared to the conventional FD approaches.
| [
{
"version": "v1",
"created": "Tue, 29 Nov 2022 06:24:14 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 06:05:59 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Moon",
"HyeonCheol",
""
],
[
"Jeong",
"JinWoo",
""
],
[
"Kim",
"SungJei",
""
]
] | TITLE: Feature-domain Adaptive Contrastive Distillation for Efficient Single
Image Super-Resolution
ABSTRACT: Recently, CNN-based SISR has numerous parameters and high computational cost
to achieve better performance, limiting its applicability to
resource-constrained devices such as mobile. As one of the methods to make the
network efficient, Knowledge Distillation (KD), which transfers teacher's
useful knowledge to student, is currently being studied. More recently, KD for
SISR utilizes Feature Distillation (FD) to minimize the Euclidean distance loss
of feature maps between teacher and student networks, but it does not
sufficiently consider how to effectively and meaningfully deliver knowledge
from teacher to improve the student performance at given network capacity
constraints. In this paper, we propose a feature-domain adaptive contrastive
distillation (FACD) method for efficiently training lightweight student SISR
networks. We show the limitations of the existing FD methods using Euclidean
distance loss, and propose a feature-domain contrastive loss that makes a
student network learn richer information from the teacher's representation in
the feature domain. In addition, we propose an adaptive distillation that
selectively applies distillation depending on the conditions of the training
patches. The experimental results show that the student EDSR and RCAN networks
with the proposed FACD scheme improves not only the PSNR performance of the
entire benchmark datasets and scales, but also the subjective image quality
compared to the conventional FD approaches.
| no_new_dataset | 0.94545 |
2212.01386 | Saurabh Deshpande Mr. | Saurabh Deshpande, Ra\'ul I. Sosa, St\'ephane P.A. Bordas, Jakub
Lengiewicz | Convolution, aggregation and attention based deep neural networks for
accelerating simulations in mechanics | null | Front. Mater. 10:1128954 | 10.3389/fmats.2023.1128954 | null | cs.LG cs.CE | http://creativecommons.org/licenses/by/4.0/ | Deep learning surrogate models are being increasingly used in accelerating
scientific simulations as a replacement for costly conventional numerical
techniques. However, their use remains a significant challenge when dealing
with real-world complex examples. In this work, we demonstrate three types of
neural network architectures for efficient learning of highly non-linear
deformations of solid bodies. The first two architectures are based on the
recently proposed CNN U-NET and MAgNET (graph U-NET) frameworks which have
shown promising performance for learning on mesh-based data. The third
architecture is Perceiver IO, a very recent architecture that belongs to the
family of attention-based neural networks--a class that has revolutionised
diverse engineering fields and is still unexplored in computational mechanics.
We study and compare the performance of all three networks on two benchmark
examples, and show their capabilities to accurately predict the non-linear
mechanical responses of soft bodies.
| [
{
"version": "v1",
"created": "Thu, 1 Dec 2022 13:10:56 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 10:41:48 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Deshpande",
"Saurabh",
""
],
[
"Sosa",
"Raúl I.",
""
],
[
"Bordas",
"Stéphane P. A.",
""
],
[
"Lengiewicz",
"Jakub",
""
]
] | TITLE: Convolution, aggregation and attention based deep neural networks for
accelerating simulations in mechanics
ABSTRACT: Deep learning surrogate models are being increasingly used in accelerating
scientific simulations as a replacement for costly conventional numerical
techniques. However, their use remains a significant challenge when dealing
with real-world complex examples. In this work, we demonstrate three types of
neural network architectures for efficient learning of highly non-linear
deformations of solid bodies. The first two architectures are based on the
recently proposed CNN U-NET and MAgNET (graph U-NET) frameworks which have
shown promising performance for learning on mesh-based data. The third
architecture is Perceiver IO, a very recent architecture that belongs to the
family of attention-based neural networks--a class that has revolutionised
diverse engineering fields and is still unexplored in computational mechanics.
We study and compare the performance of all three networks on two benchmark
examples, and show their capabilities to accurately predict the non-linear
mechanical responses of soft bodies.
| no_new_dataset | 0.947478 |
2212.02951 | Ziqi Wang | Ziqi Wang, Tianye Shu, Jialin Liu | State Space Closure: Revisiting Endless Online Level Generation via
Reinforcement Learning | Accepted by the IEEE Transactions on Games | null | null | null | cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this paper, we revisit endless online level generation with the recently
proposed experience-driven procedural content generation via reinforcement
learning (EDRL) framework. Inspired by an observation that EDRL tends to
generate recurrent patterns, we formulate a notion of state space closure which
makes any stochastic state appeared possibly in an infinite-horizon online
generation process can be found within a finite-horizon. Through theoretical
analysis, we find that even though state space closure arises a concern about
diversity, it generalises EDRL trained with a finite-horizon to the
infinite-horizon scenario without deterioration of content quality. Moreover,
we verify the quality and the diversity of contents generated by EDRL via
empirical studies, on the widely used Super Mario Bros. benchmark. Experimental
results reveal that the diversity of levels generated by EDRL is limited due to
the state space closure, whereas their quality does not deteriorate in a
horizon which is longer than the one specified in the training. Concluding our
outcomes and analysis, future work on endless online level generation via
reinforcement learning should address the issue of diversity while assuring the
occurrence of state space closure and quality.
| [
{
"version": "v1",
"created": "Tue, 6 Dec 2022 13:12:16 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 10:23:22 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Wang",
"Ziqi",
""
],
[
"Shu",
"Tianye",
""
],
[
"Liu",
"Jialin",
""
]
] | TITLE: State Space Closure: Revisiting Endless Online Level Generation via
Reinforcement Learning
ABSTRACT: In this paper, we revisit endless online level generation with the recently
proposed experience-driven procedural content generation via reinforcement
learning (EDRL) framework. Inspired by an observation that EDRL tends to
generate recurrent patterns, we formulate a notion of state space closure which
makes any stochastic state appeared possibly in an infinite-horizon online
generation process can be found within a finite-horizon. Through theoretical
analysis, we find that even though state space closure arises a concern about
diversity, it generalises EDRL trained with a finite-horizon to the
infinite-horizon scenario without deterioration of content quality. Moreover,
we verify the quality and the diversity of contents generated by EDRL via
empirical studies, on the widely used Super Mario Bros. benchmark. Experimental
results reveal that the diversity of levels generated by EDRL is limited due to
the state space closure, whereas their quality does not deteriorate in a
horizon which is longer than the one specified in the training. Concluding our
outcomes and analysis, future work on endless online level generation via
reinforcement learning should address the issue of diversity while assuring the
occurrence of state space closure and quality.
| no_new_dataset | 0.954095 |
2212.06202 | Chengzhi Mao | Chengzhi Mao, Revant Teotia, Amrutha Sundar, Sachit Menon, Junfeng
Yang, Xin Wang, Carl Vondrick | Doubly Right Object Recognition: A Why Prompt for Visual Rationales | Accepted at CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Many visual recognition models are evaluated only on their classification
accuracy, a metric for which they obtain strong performance. In this paper, we
investigate whether computer vision models can also provide correct rationales
for their predictions. We propose a ``doubly right'' object recognition
benchmark, where the metric requires the model to simultaneously produce both
the right labels as well as the right rationales. We find that state-of-the-art
visual models, such as CLIP, often provide incorrect rationales for their
categorical predictions. However, by transferring the rationales from language
models into visual representations through a tailored dataset, we show that we
can learn a ``why prompt,'' which adapts large visual representations to
produce correct rationales. Visualizations and empirical experiments show that
our prompts significantly improve performance on doubly right object
recognition, in addition to zero-shot transfer to unseen tasks and datasets.
| [
{
"version": "v1",
"created": "Mon, 12 Dec 2022 19:25:45 GMT"
},
{
"version": "v2",
"created": "Wed, 22 Mar 2023 21:05:49 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Mao",
"Chengzhi",
""
],
[
"Teotia",
"Revant",
""
],
[
"Sundar",
"Amrutha",
""
],
[
"Menon",
"Sachit",
""
],
[
"Yang",
"Junfeng",
""
],
[
"Wang",
"Xin",
""
],
[
"Vondrick",
"Carl",
""
]
] | TITLE: Doubly Right Object Recognition: A Why Prompt for Visual Rationales
ABSTRACT: Many visual recognition models are evaluated only on their classification
accuracy, a metric for which they obtain strong performance. In this paper, we
investigate whether computer vision models can also provide correct rationales
for their predictions. We propose a ``doubly right'' object recognition
benchmark, where the metric requires the model to simultaneously produce both
the right labels as well as the right rationales. We find that state-of-the-art
visual models, such as CLIP, often provide incorrect rationales for their
categorical predictions. However, by transferring the rationales from language
models into visual representations through a tailored dataset, we show that we
can learn a ``why prompt,'' which adapts large visual representations to
produce correct rationales. Visualizations and empirical experiments show that
our prompts significantly improve performance on doubly right object
recognition, in addition to zero-shot transfer to unseen tasks and datasets.
| no_new_dataset | 0.72628 |
2212.09478 | Ludan Ruan | Ludan Ruan and Yiyang Ma and Huan Yang and Huiguo He and Bei Liu and
Jianlong Fu and Nicholas Jing Yuan and Qin Jin and Baining Guo | MM-Diffusion: Learning Multi-Modal Diffusion Models for Joint Audio and
Video Generation | Accepted by CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | We propose the first joint audio-video generation framework that brings
engaging watching and listening experiences simultaneously, towards
high-quality realistic videos. To generate joint audio-video pairs, we propose
a novel Multi-Modal Diffusion model (i.e., MM-Diffusion), with two-coupled
denoising autoencoders. In contrast to existing single-modal diffusion models,
MM-Diffusion consists of a sequential multi-modal U-Net for a joint denoising
process by design. Two subnets for audio and video learn to gradually generate
aligned audio-video pairs from Gaussian noises. To ensure semantic consistency
across modalities, we propose a novel random-shift based attention block
bridging over the two subnets, which enables efficient cross-modal alignment,
and thus reinforces the audio-video fidelity for each other. Extensive
experiments show superior results in unconditional audio-video generation, and
zero-shot conditional tasks (e.g., video-to-audio). In particular, we achieve
the best FVD and FAD on Landscape and AIST++ dancing datasets. Turing tests of
10k votes further demonstrate dominant preferences for our model. The code and
pre-trained models can be downloaded at
https://github.com/researchmm/MM-Diffusion.
| [
{
"version": "v1",
"created": "Mon, 19 Dec 2022 14:11:52 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 12:13:52 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Ruan",
"Ludan",
""
],
[
"Ma",
"Yiyang",
""
],
[
"Yang",
"Huan",
""
],
[
"He",
"Huiguo",
""
],
[
"Liu",
"Bei",
""
],
[
"Fu",
"Jianlong",
""
],
[
"Yuan",
"Nicholas Jing",
""
],
[
"Jin",
"Qin",
""
],
[
"Guo",
"Baining",
""
]
] | TITLE: MM-Diffusion: Learning Multi-Modal Diffusion Models for Joint Audio and
Video Generation
ABSTRACT: We propose the first joint audio-video generation framework that brings
engaging watching and listening experiences simultaneously, towards
high-quality realistic videos. To generate joint audio-video pairs, we propose
a novel Multi-Modal Diffusion model (i.e., MM-Diffusion), with two-coupled
denoising autoencoders. In contrast to existing single-modal diffusion models,
MM-Diffusion consists of a sequential multi-modal U-Net for a joint denoising
process by design. Two subnets for audio and video learn to gradually generate
aligned audio-video pairs from Gaussian noises. To ensure semantic consistency
across modalities, we propose a novel random-shift based attention block
bridging over the two subnets, which enables efficient cross-modal alignment,
and thus reinforces the audio-video fidelity for each other. Extensive
experiments show superior results in unconditional audio-video generation, and
zero-shot conditional tasks (e.g., video-to-audio). In particular, we achieve
the best FVD and FAD on Landscape and AIST++ dancing datasets. Turing tests of
10k votes further demonstrate dominant preferences for our model. The code and
pre-trained models can be downloaded at
https://github.com/researchmm/MM-Diffusion.
| no_new_dataset | 0.949716 |
2212.10806 | Jongbeom Baek | Jongbeom Baek, Gyeongnyeon Kim, Seonghoon Park, Honggyu An, Matteo
Poggi, Seungryong Kim | MaskingDepth: Masked Consistency Regularization for Semi-supervised
Monocular Depth Estimation | Project page: https://ku-cvlab.github.io/MaskingDepth/ | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | We propose MaskingDepth, a novel semi-supervised learning framework for
monocular depth estimation to mitigate the reliance on large ground-truth depth
quantities. MaskingDepth is designed to enforce consistency between the
strongly-augmented unlabeled data and the pseudo-labels derived from
weakly-augmented unlabeled data, which enables learning depth without
supervision. In this framework, a novel data augmentation is proposed to take
the advantage of a naive masking strategy as an augmentation, while avoiding
its scale ambiguity problem between depths from weakly- and strongly-augmented
branches and risk of missing small-scale instances. To only retain
high-confident depth predictions from the weakly-augmented branch as
pseudo-labels, we also present an uncertainty estimation technique, which is
used to define robust consistency regularization. Experiments on KITTI and
NYU-Depth-v2 datasets demonstrate the effectiveness of each component, its
robustness to the use of fewer depth-annotated images, and superior performance
compared to other state-of-the-art semi-supervised methods for monocular depth
estimation. Furthermore, we show our method can be easily extended to domain
adaptation task. Our code is available at
https://github.com/KU-CVLAB/MaskingDepth.
| [
{
"version": "v1",
"created": "Wed, 21 Dec 2022 06:56:22 GMT"
},
{
"version": "v2",
"created": "Thu, 22 Dec 2022 02:37:48 GMT"
},
{
"version": "v3",
"created": "Thu, 23 Mar 2023 23:05:20 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Baek",
"Jongbeom",
""
],
[
"Kim",
"Gyeongnyeon",
""
],
[
"Park",
"Seonghoon",
""
],
[
"An",
"Honggyu",
""
],
[
"Poggi",
"Matteo",
""
],
[
"Kim",
"Seungryong",
""
]
] | TITLE: MaskingDepth: Masked Consistency Regularization for Semi-supervised
Monocular Depth Estimation
ABSTRACT: We propose MaskingDepth, a novel semi-supervised learning framework for
monocular depth estimation to mitigate the reliance on large ground-truth depth
quantities. MaskingDepth is designed to enforce consistency between the
strongly-augmented unlabeled data and the pseudo-labels derived from
weakly-augmented unlabeled data, which enables learning depth without
supervision. In this framework, a novel data augmentation is proposed to take
the advantage of a naive masking strategy as an augmentation, while avoiding
its scale ambiguity problem between depths from weakly- and strongly-augmented
branches and risk of missing small-scale instances. To only retain
high-confident depth predictions from the weakly-augmented branch as
pseudo-labels, we also present an uncertainty estimation technique, which is
used to define robust consistency regularization. Experiments on KITTI and
NYU-Depth-v2 datasets demonstrate the effectiveness of each component, its
robustness to the use of fewer depth-annotated images, and superior performance
compared to other state-of-the-art semi-supervised methods for monocular depth
estimation. Furthermore, we show our method can be easily extended to domain
adaptation task. Our code is available at
https://github.com/KU-CVLAB/MaskingDepth.
| no_new_dataset | 0.946745 |
2212.11934 | Margarita Chasapi | Margarita Chasapi, Pablo Antolin, Annalisa Buffa | A localized reduced basis approach for unfitted domain methods on
parameterized geometries | 35 pages, 19 figures, 5 tables. Preprint accepted in Computer Methods
in Applied Mechanics and Engineering | Computer Methods in Applied Mechanics and Engineering 2023.115997 | 10.1016/j.cma.2023.115997 | null | math.NA cs.NA | http://creativecommons.org/licenses/by-nc-nd/4.0/ | This work introduces a reduced order modeling (ROM) framework for the
solution of parameterized second-order linear elliptic partial differential
equations formulated on unfitted geometries. The goal is to construct efficient
projection-based ROMs, which rely on techniques such as the reduced basis
method and discrete empirical interpolation. The presence of geometrical
parameters in unfitted domain discretizations entails challenges for the
application of standard ROMs. Therefore, in this work we propose a methodology
based on i) extension of snapshots on the background mesh and ii) localization
strategies to decrease the number of reduced basis functions. The method we
obtain is computationally efficient and accurate, while it is agnostic with
respect to the underlying discretization choice. We test the applicability of
the proposed framework with numerical experiments on two model problems, namely
the Poisson and linear elasticity problems. In particular, we study several
benchmarks formulated on two-dimensional, trimmed domains discretized with
splines and we observe a significant reduction of the online computational cost
compared to standard ROMs for the same level of accuracy. Moreover, we show the
applicability of our methodology to a three-dimensional geometry of a linear
elastic problem.
| [
{
"version": "v1",
"created": "Thu, 22 Dec 2022 18:14:34 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 17:30:55 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Chasapi",
"Margarita",
""
],
[
"Antolin",
"Pablo",
""
],
[
"Buffa",
"Annalisa",
""
]
] | TITLE: A localized reduced basis approach for unfitted domain methods on
parameterized geometries
ABSTRACT: This work introduces a reduced order modeling (ROM) framework for the
solution of parameterized second-order linear elliptic partial differential
equations formulated on unfitted geometries. The goal is to construct efficient
projection-based ROMs, which rely on techniques such as the reduced basis
method and discrete empirical interpolation. The presence of geometrical
parameters in unfitted domain discretizations entails challenges for the
application of standard ROMs. Therefore, in this work we propose a methodology
based on i) extension of snapshots on the background mesh and ii) localization
strategies to decrease the number of reduced basis functions. The method we
obtain is computationally efficient and accurate, while it is agnostic with
respect to the underlying discretization choice. We test the applicability of
the proposed framework with numerical experiments on two model problems, namely
the Poisson and linear elasticity problems. In particular, we study several
benchmarks formulated on two-dimensional, trimmed domains discretized with
splines and we observe a significant reduction of the online computational cost
compared to standard ROMs for the same level of accuracy. Moreover, we show the
applicability of our methodology to a three-dimensional geometry of a linear
elastic problem.
| no_new_dataset | 0.947284 |
2212.12402 | Shenglan Du | Shenglan Du, Nail Ibrahimli, Jantien Stoter, Julian Kooij, Liangliang
Nan | Push-the-Boundary: Boundary-aware Feature Propagation for Semantic
Segmentation of 3D Point Clouds | 3DV2022 | null | 10.1109/3DV57658.2022.00025 | null | cs.CV | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Feedforward fully convolutional neural networks currently dominate in
semantic segmentation of 3D point clouds. Despite their great success, they
suffer from the loss of local information at low-level layers, posing
significant challenges to accurate scene segmentation and precise object
boundary delineation. Prior works either address this issue by post-processing
or jointly learn object boundaries to implicitly improve feature encoding of
the networks. These approaches often require additional modules which are
difficult to integrate into the original architecture.
To improve the segmentation near object boundaries, we propose a
boundary-aware feature propagation mechanism. This mechanism is achieved by
exploiting a multi-task learning framework that aims to explicitly guide the
boundaries to their original locations. With one shared encoder, our network
outputs (i) boundary localization, (ii) prediction of directions pointing to
the object's interior, and (iii) semantic segmentation, in three parallel
streams. The predicted boundaries and directions are fused to propagate the
learned features to refine the segmentation. We conduct extensive experiments
on the S3DIS and SensatUrban datasets against various baseline methods,
demonstrating that our proposed approach yields consistent improvements by
reducing boundary errors. Our code is available at
https://github.com/shenglandu/PushBoundary.
| [
{
"version": "v1",
"created": "Fri, 23 Dec 2022 15:42:01 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Du",
"Shenglan",
""
],
[
"Ibrahimli",
"Nail",
""
],
[
"Stoter",
"Jantien",
""
],
[
"Kooij",
"Julian",
""
],
[
"Nan",
"Liangliang",
""
]
] | TITLE: Push-the-Boundary: Boundary-aware Feature Propagation for Semantic
Segmentation of 3D Point Clouds
ABSTRACT: Feedforward fully convolutional neural networks currently dominate in
semantic segmentation of 3D point clouds. Despite their great success, they
suffer from the loss of local information at low-level layers, posing
significant challenges to accurate scene segmentation and precise object
boundary delineation. Prior works either address this issue by post-processing
or jointly learn object boundaries to implicitly improve feature encoding of
the networks. These approaches often require additional modules which are
difficult to integrate into the original architecture.
To improve the segmentation near object boundaries, we propose a
boundary-aware feature propagation mechanism. This mechanism is achieved by
exploiting a multi-task learning framework that aims to explicitly guide the
boundaries to their original locations. With one shared encoder, our network
outputs (i) boundary localization, (ii) prediction of directions pointing to
the object's interior, and (iii) semantic segmentation, in three parallel
streams. The predicted boundaries and directions are fused to propagate the
learned features to refine the segmentation. We conduct extensive experiments
on the S3DIS and SensatUrban datasets against various baseline methods,
demonstrating that our proposed approach yields consistent improvements by
reducing boundary errors. Our code is available at
https://github.com/shenglandu/PushBoundary.
| no_new_dataset | 0.949435 |
2212.14268 | Krystian Radlak | Bartlomiej Olber, Krystian Radlak, Adam Popowicz, Michal
Szczepankiewicz, Krystian Chachu{\l}a | Detection of out-of-distribution samples using binary neuron activation
patterns | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by-nc-sa/4.0/ | Deep neural networks (DNN) have outstanding performance in various
applications. Despite numerous efforts of the research community,
out-of-distribution (OOD) samples remain a significant limitation of DNN
classifiers. The ability to identify previously unseen inputs as novel is
crucial in safety-critical applications such as self-driving cars, unmanned
aerial vehicles, and robots. Existing approaches to detect OOD samples treat a
DNN as a black box and evaluate the confidence score of the output predictions.
Unfortunately, this method frequently fails, because DNNs are not trained to
reduce their confidence for OOD inputs. In this work, we introduce a novel
method for OOD detection. Our method is motivated by theoretical analysis of
neuron activation patterns (NAP) in ReLU-based architectures. The proposed
method does not introduce a high computational overhead due to the binary
representation of the activation patterns extracted from convolutional layers.
The extensive empirical evaluation proves its high performance on various DNN
architectures and seven image datasets.
| [
{
"version": "v1",
"created": "Thu, 29 Dec 2022 11:42:46 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 16:10:35 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Olber",
"Bartlomiej",
""
],
[
"Radlak",
"Krystian",
""
],
[
"Popowicz",
"Adam",
""
],
[
"Szczepankiewicz",
"Michal",
""
],
[
"Chachuła",
"Krystian",
""
]
] | TITLE: Detection of out-of-distribution samples using binary neuron activation
patterns
ABSTRACT: Deep neural networks (DNN) have outstanding performance in various
applications. Despite numerous efforts of the research community,
out-of-distribution (OOD) samples remain a significant limitation of DNN
classifiers. The ability to identify previously unseen inputs as novel is
crucial in safety-critical applications such as self-driving cars, unmanned
aerial vehicles, and robots. Existing approaches to detect OOD samples treat a
DNN as a black box and evaluate the confidence score of the output predictions.
Unfortunately, this method frequently fails, because DNNs are not trained to
reduce their confidence for OOD inputs. In this work, we introduce a novel
method for OOD detection. Our method is motivated by theoretical analysis of
neuron activation patterns (NAP) in ReLU-based architectures. The proposed
method does not introduce a high computational overhead due to the binary
representation of the activation patterns extracted from convolutional layers.
The extensive empirical evaluation proves its high performance on various DNN
architectures and seven image datasets.
| no_new_dataset | 0.944125 |
2301.04467 | Yuqi Wang | Yuqi Wang, Yuntao Chen, and Zhaoxiang Zhang | FrustumFormer: Adaptive Instance-aware Resampling for Multi-view 3D
Detection | Accepted to CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | The transformation of features from 2D perspective space to 3D space is
essential to multi-view 3D object detection. Recent approaches mainly focus on
the design of view transformation, either pixel-wisely lifting perspective view
features into 3D space with estimated depth or grid-wisely constructing BEV
features via 3D projection, treating all pixels or grids equally. However,
choosing what to transform is also important but has rarely been discussed
before. The pixels of a moving car are more informative than the pixels of the
sky. To fully utilize the information contained in images, the view
transformation should be able to adapt to different image regions according to
their contents. In this paper, we propose a novel framework named
FrustumFormer, which pays more attention to the features in instance regions
via adaptive instance-aware resampling. Specifically, the model obtains
instance frustums on the bird's eye view by leveraging image view object
proposals. An adaptive occupancy mask within the instance frustum is learned to
refine the instance location. Moreover, the temporal frustum intersection could
further reduce the localization uncertainty of objects. Comprehensive
experiments on the nuScenes dataset demonstrate the effectiveness of
FrustumFormer, and we achieve a new state-of-the-art performance on the
benchmark. Codes and models will be made available at
https://github.com/Robertwyq/Frustum.
| [
{
"version": "v1",
"created": "Tue, 10 Jan 2023 17:51:55 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 09:18:39 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Wang",
"Yuqi",
""
],
[
"Chen",
"Yuntao",
""
],
[
"Zhang",
"Zhaoxiang",
""
]
] | TITLE: FrustumFormer: Adaptive Instance-aware Resampling for Multi-view 3D
Detection
ABSTRACT: The transformation of features from 2D perspective space to 3D space is
essential to multi-view 3D object detection. Recent approaches mainly focus on
the design of view transformation, either pixel-wisely lifting perspective view
features into 3D space with estimated depth or grid-wisely constructing BEV
features via 3D projection, treating all pixels or grids equally. However,
choosing what to transform is also important but has rarely been discussed
before. The pixels of a moving car are more informative than the pixels of the
sky. To fully utilize the information contained in images, the view
transformation should be able to adapt to different image regions according to
their contents. In this paper, we propose a novel framework named
FrustumFormer, which pays more attention to the features in instance regions
via adaptive instance-aware resampling. Specifically, the model obtains
instance frustums on the bird's eye view by leveraging image view object
proposals. An adaptive occupancy mask within the instance frustum is learned to
refine the instance location. Moreover, the temporal frustum intersection could
further reduce the localization uncertainty of objects. Comprehensive
experiments on the nuScenes dataset demonstrate the effectiveness of
FrustumFormer, and we achieve a new state-of-the-art performance on the
benchmark. Codes and models will be made available at
https://github.com/Robertwyq/Frustum.
| no_new_dataset | 0.949902 |
2301.05709 | Anas Mahmoud | Anas Mahmoud, Jordan S. K. Hu, Tianshu Kuai, Ali Harakeh, Liam Paull,
and Steven L. Waslander | Self-Supervised Image-to-Point Distillation via Semantically Tolerant
Contrastive Loss | Accepted in CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | An effective framework for learning 3D representations for perception tasks
is distilling rich self-supervised image features via contrastive learning.
However, image-to point representation learning for autonomous driving datasets
faces two main challenges: 1) the abundance of self-similarity, which results
in the contrastive losses pushing away semantically similar point and image
regions and thus disturbing the local semantic structure of the learned
representations, and 2) severe class imbalance as pretraining gets dominated by
over-represented classes. We propose to alleviate the self-similarity problem
through a novel semantically tolerant image-to-point contrastive loss that
takes into consideration the semantic distance between positive and negative
image regions to minimize contrasting semantically similar point and image
regions. Additionally, we address class imbalance by designing a class-agnostic
balanced loss that approximates the degree of class imbalance through an
aggregate sample-to-samples semantic similarity measure. We demonstrate that
our semantically-tolerant contrastive loss with class balancing improves
state-of-the art 2D-to-3D representation learning in all evaluation settings on
3D semantic segmentation. Our method consistently outperforms state-of-the-art
2D-to-3D representation learning frameworks across a wide range of 2D
self-supervised pretrained models.
| [
{
"version": "v1",
"created": "Thu, 12 Jan 2023 19:58:54 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 15:53:21 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Mahmoud",
"Anas",
""
],
[
"Hu",
"Jordan S. K.",
""
],
[
"Kuai",
"Tianshu",
""
],
[
"Harakeh",
"Ali",
""
],
[
"Paull",
"Liam",
""
],
[
"Waslander",
"Steven L.",
""
]
] | TITLE: Self-Supervised Image-to-Point Distillation via Semantically Tolerant
Contrastive Loss
ABSTRACT: An effective framework for learning 3D representations for perception tasks
is distilling rich self-supervised image features via contrastive learning.
However, image-to point representation learning for autonomous driving datasets
faces two main challenges: 1) the abundance of self-similarity, which results
in the contrastive losses pushing away semantically similar point and image
regions and thus disturbing the local semantic structure of the learned
representations, and 2) severe class imbalance as pretraining gets dominated by
over-represented classes. We propose to alleviate the self-similarity problem
through a novel semantically tolerant image-to-point contrastive loss that
takes into consideration the semantic distance between positive and negative
image regions to minimize contrasting semantically similar point and image
regions. Additionally, we address class imbalance by designing a class-agnostic
balanced loss that approximates the degree of class imbalance through an
aggregate sample-to-samples semantic similarity measure. We demonstrate that
our semantically-tolerant contrastive loss with class balancing improves
state-of-the art 2D-to-3D representation learning in all evaluation settings on
3D semantic segmentation. Our method consistently outperforms state-of-the-art
2D-to-3D representation learning frameworks across a wide range of 2D
self-supervised pretrained models.
| no_new_dataset | 0.948251 |
2301.13760 | Pascal Nasahl | Pascal Nasahl, Salmin Sultana, Hans Liljestrand, Karanvir Grewal,
Michael LeMay, David M. Durham, David Schrammel, Stefan Mangard | EC-CFI: Control-Flow Integrity via Code Encryption Counteracting Fault
Attacks | Accepted at HOST'23 | null | null | null | cs.CR | http://creativecommons.org/licenses/by/4.0/ | Fault attacks enable adversaries to manipulate the control-flow of
security-critical applications. By inducing targeted faults into the CPU, the
software's call graph can be escaped and the control-flow can be redirected to
arbitrary functions inside the program. To protect the control-flow from these
attacks, dedicated fault control-flow integrity (CFI) countermeasures are
commonly deployed. However, these schemes either have high detection latencies
or require intrusive hardware changes. In this paper, we present EC-CFI, a
software-based cryptographically enforced CFI scheme with no detection latency
utilizing hardware features of recent Intel platforms. Our EC-CFI prototype is
designed to prevent an adversary from escaping the program's call graph using
faults by encrypting each function with a different key before execution. At
runtime, the instrumented program dynamically derives the decryption key,
ensuring that the code only can be successfully decrypted when the program
follows the intended call graph. To enable this level of protection on Intel
commodity systems, we introduce extended page table (EPT) aliasing allowing us
to achieve function-granular encryption by combing Intel's TME-MK and
virtualization technology. We open-source our custom LLVM-based toolchain
automatically protecting arbitrary programs with EC-CFI. Furthermore, we
evaluate our EPT aliasing approach with the SPEC CPU2017 and Embench-IoT
benchmarks and discuss and evaluate potential TME-MK hardware changes
minimizing runtime overheads.
| [
{
"version": "v1",
"created": "Tue, 31 Jan 2023 16:51:33 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 10:41:21 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Nasahl",
"Pascal",
""
],
[
"Sultana",
"Salmin",
""
],
[
"Liljestrand",
"Hans",
""
],
[
"Grewal",
"Karanvir",
""
],
[
"LeMay",
"Michael",
""
],
[
"Durham",
"David M.",
""
],
[
"Schrammel",
"David",
""
],
[
"Mangard",
"Stefan",
""
]
] | TITLE: EC-CFI: Control-Flow Integrity via Code Encryption Counteracting Fault
Attacks
ABSTRACT: Fault attacks enable adversaries to manipulate the control-flow of
security-critical applications. By inducing targeted faults into the CPU, the
software's call graph can be escaped and the control-flow can be redirected to
arbitrary functions inside the program. To protect the control-flow from these
attacks, dedicated fault control-flow integrity (CFI) countermeasures are
commonly deployed. However, these schemes either have high detection latencies
or require intrusive hardware changes. In this paper, we present EC-CFI, a
software-based cryptographically enforced CFI scheme with no detection latency
utilizing hardware features of recent Intel platforms. Our EC-CFI prototype is
designed to prevent an adversary from escaping the program's call graph using
faults by encrypting each function with a different key before execution. At
runtime, the instrumented program dynamically derives the decryption key,
ensuring that the code only can be successfully decrypted when the program
follows the intended call graph. To enable this level of protection on Intel
commodity systems, we introduce extended page table (EPT) aliasing allowing us
to achieve function-granular encryption by combing Intel's TME-MK and
virtualization technology. We open-source our custom LLVM-based toolchain
automatically protecting arbitrary programs with EC-CFI. Furthermore, we
evaluate our EPT aliasing approach with the SPEC CPU2017 and Embench-IoT
benchmarks and discuss and evaluate potential TME-MK hardware changes
minimizing runtime overheads.
| no_new_dataset | 0.940188 |
2302.01047 | Yasir Ghunaim | Yasir Ghunaim, Adel Bibi, Kumail Alhamoud, Motasem Alfarra, Hasan Abed
Al Kader Hammoud, Ameya Prabhu, Philip H. S. Torr, Bernard Ghanem | Real-Time Evaluation in Online Continual Learning: A New Hope | Accepted at CVPR'23 as Highlight (Top 2.5%) | null | null | null | cs.LG cs.AI cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Current evaluations of Continual Learning (CL) methods typically assume that
there is no constraint on training time and computation. This is an unrealistic
assumption for any real-world setting, which motivates us to propose: a
practical real-time evaluation of continual learning, in which the stream does
not wait for the model to complete training before revealing the next data for
predictions. To do this, we evaluate current CL methods with respect to their
computational costs. We conduct extensive experiments on CLOC, a large-scale
dataset containing 39 million time-stamped images with geolocation labels. We
show that a simple baseline outperforms state-of-the-art CL methods under this
evaluation, questioning the applicability of existing methods in realistic
settings. In addition, we explore various CL components commonly used in the
literature, including memory sampling strategies and regularization approaches.
We find that all considered methods fail to be competitive against our simple
baseline. This surprisingly suggests that the majority of existing CL
literature is tailored to a specific class of streams that is not practical. We
hope that the evaluation we provide will be the first step towards a paradigm
shift to consider the computational cost in the development of online continual
learning methods.
| [
{
"version": "v1",
"created": "Thu, 2 Feb 2023 12:21:10 GMT"
},
{
"version": "v2",
"created": "Thu, 23 Mar 2023 12:30:35 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 07:23:36 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Ghunaim",
"Yasir",
""
],
[
"Bibi",
"Adel",
""
],
[
"Alhamoud",
"Kumail",
""
],
[
"Alfarra",
"Motasem",
""
],
[
"Hammoud",
"Hasan Abed Al Kader",
""
],
[
"Prabhu",
"Ameya",
""
],
[
"Torr",
"Philip H. S.",
""
],
[
"Ghanem",
"Bernard",
""
]
] | TITLE: Real-Time Evaluation in Online Continual Learning: A New Hope
ABSTRACT: Current evaluations of Continual Learning (CL) methods typically assume that
there is no constraint on training time and computation. This is an unrealistic
assumption for any real-world setting, which motivates us to propose: a
practical real-time evaluation of continual learning, in which the stream does
not wait for the model to complete training before revealing the next data for
predictions. To do this, we evaluate current CL methods with respect to their
computational costs. We conduct extensive experiments on CLOC, a large-scale
dataset containing 39 million time-stamped images with geolocation labels. We
show that a simple baseline outperforms state-of-the-art CL methods under this
evaluation, questioning the applicability of existing methods in realistic
settings. In addition, we explore various CL components commonly used in the
literature, including memory sampling strategies and regularization approaches.
We find that all considered methods fail to be competitive against our simple
baseline. This surprisingly suggests that the majority of existing CL
literature is tailored to a specific class of streams that is not practical. We
hope that the evaluation we provide will be the first step towards a paradigm
shift to consider the computational cost in the development of online continual
learning methods.
| no_new_dataset | 0.925432 |
2302.01574 | Brian Hsu | Brian Hsu, Xiaotong Chen, Ying Han, Hongseok Namkoong, Kinjal Basu | An Operational Perspective to Fairness Interventions: Where and How to
Intervene | null | null | null | null | cs.LG | http://creativecommons.org/licenses/by/4.0/ | As AI-based decision systems proliferate, their successful operationalization
requires balancing multiple desiderata: predictive performance, disparity
across groups, safeguarding sensitive group attributes (e.g., race), and
engineering cost. We present a holistic framework for evaluating and
contextualizing fairness interventions with respect to the above desiderata.
The two key points of practical consideration are \emph{where} (pre-, in-,
post-processing) and \emph{how} (in what way the sensitive group data is used)
the intervention is introduced. We demonstrate our framework with a case study
on predictive parity. In it, we first propose a novel method for achieving
predictive parity fairness without using group data at inference time via
distibutionally robust optimization. Then, we showcase the effectiveness of
these methods in a benchmarking study of close to 400 variations across two
major model types (XGBoost vs. Neural Net), ten datasets, and over twenty
unique methodologies. Methodological insights derived from our empirical study
inform the practical design of ML workflow with fairness as a central concern.
We find predictive parity is difficult to achieve without using group data, and
despite requiring group data during model training (but not inference),
distributionally robust methods we develop provide significant Pareto
improvement. Moreover, a plain XGBoost model often Pareto-dominates neural
networks with fairness interventions, highlighting the importance of model
inductive bias.
| [
{
"version": "v1",
"created": "Fri, 3 Feb 2023 07:04:33 GMT"
},
{
"version": "v2",
"created": "Thu, 23 Mar 2023 21:20:38 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Hsu",
"Brian",
""
],
[
"Chen",
"Xiaotong",
""
],
[
"Han",
"Ying",
""
],
[
"Namkoong",
"Hongseok",
""
],
[
"Basu",
"Kinjal",
""
]
] | TITLE: An Operational Perspective to Fairness Interventions: Where and How to
Intervene
ABSTRACT: As AI-based decision systems proliferate, their successful operationalization
requires balancing multiple desiderata: predictive performance, disparity
across groups, safeguarding sensitive group attributes (e.g., race), and
engineering cost. We present a holistic framework for evaluating and
contextualizing fairness interventions with respect to the above desiderata.
The two key points of practical consideration are \emph{where} (pre-, in-,
post-processing) and \emph{how} (in what way the sensitive group data is used)
the intervention is introduced. We demonstrate our framework with a case study
on predictive parity. In it, we first propose a novel method for achieving
predictive parity fairness without using group data at inference time via
distibutionally robust optimization. Then, we showcase the effectiveness of
these methods in a benchmarking study of close to 400 variations across two
major model types (XGBoost vs. Neural Net), ten datasets, and over twenty
unique methodologies. Methodological insights derived from our empirical study
inform the practical design of ML workflow with fairness as a central concern.
We find predictive parity is difficult to achieve without using group data, and
despite requiring group data during model training (but not inference),
distributionally robust methods we develop provide significant Pareto
improvement. Moreover, a plain XGBoost model often Pareto-dominates neural
networks with fairness interventions, highlighting the importance of model
inductive bias.
| no_new_dataset | 0.947624 |
2302.06418 | Ricardo Macedo | Ricardo Macedo, Mariana Miranda, Yusuke Tanimura, Jason Haga, Amit
Ruhela, Stephen Lien Harrell, Richard Todd Evans, Jos\'e Pereira, Jo\~ao
Paulo | PADLL: Taming Metadata-intensive HPC Jobs Through Dynamic,
Application-agnostic QoS Control | To appear at 23rd IEEE/ACM International Symposium on Cluster, Cloud
and Internet Computing (CCGrid'23) | null | null | null | cs.DC cs.OS cs.PF | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Modern I/O applications that run on HPC infrastructures are increasingly
becoming read and metadata intensive. However, having multiple concurrent
applications submitting large amounts of metadata operations can easily
saturate the shared parallel file system's metadata resources, leading to
overall performance degradation and I/O unfairness. We present PADLL, an
application and file system agnostic storage middleware that enables QoS
control of data and metadata workflows in HPC storage systems. It adopts ideas
from Software-Defined Storage, building data plane stages that mediate and rate
limit POSIX requests submitted to the shared file system, and a control plane
that holistically coordinates how all I/O workflows are handled. We demonstrate
its performance and feasibility under multiple QoS policies using synthetic
benchmarks, real-world applications, and traces collected from a production
file system. Results show that PADLL can enforce complex storage QoS policies
over concurrent metadata-aggressive jobs, ensuring fairness and prioritization.
| [
{
"version": "v1",
"created": "Mon, 13 Feb 2023 14:55:06 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 00:25:31 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Macedo",
"Ricardo",
""
],
[
"Miranda",
"Mariana",
""
],
[
"Tanimura",
"Yusuke",
""
],
[
"Haga",
"Jason",
""
],
[
"Ruhela",
"Amit",
""
],
[
"Harrell",
"Stephen Lien",
""
],
[
"Evans",
"Richard Todd",
""
],
[
"Pereira",
"José",
""
],
[
"Paulo",
"João",
""
]
] | TITLE: PADLL: Taming Metadata-intensive HPC Jobs Through Dynamic,
Application-agnostic QoS Control
ABSTRACT: Modern I/O applications that run on HPC infrastructures are increasingly
becoming read and metadata intensive. However, having multiple concurrent
applications submitting large amounts of metadata operations can easily
saturate the shared parallel file system's metadata resources, leading to
overall performance degradation and I/O unfairness. We present PADLL, an
application and file system agnostic storage middleware that enables QoS
control of data and metadata workflows in HPC storage systems. It adopts ideas
from Software-Defined Storage, building data plane stages that mediate and rate
limit POSIX requests submitted to the shared file system, and a control plane
that holistically coordinates how all I/O workflows are handled. We demonstrate
its performance and feasibility under multiple QoS policies using synthetic
benchmarks, real-world applications, and traces collected from a production
file system. Results show that PADLL can enforce complex storage QoS policies
over concurrent metadata-aggressive jobs, ensuring fairness and prioritization.
| no_new_dataset | 0.944893 |
2302.10927 | Peichao Li | Peichao Li, Muhammad Asad, Conor Horgan, Oscar MacCormac, Jonathan
Shapey, Tom Vercauteren | Spatial gradient consistency for unsupervised learning of hyperspectral
demosaicking: Application to surgical imaging | null | International Journal of Computer Assisted Radiology and Surgery,
2023 | 10.1007/s11548-023-02865-7 | null | eess.IV cs.CV cs.LG | http://creativecommons.org/licenses/by/4.0/ | Hyperspectral imaging has the potential to improve intraoperative decision
making if tissue characterisation is performed in real-time and with
high-resolution. Hyperspectral snapshot mosaic sensors offer a promising
approach due to their fast acquisition speed and compact size. However, a
demosaicking algorithm is required to fully recover the spatial and spectral
information of the snapshot images. Most state-of-the-art demosaicking
algorithms require ground-truth training data with paired snapshot and
high-resolution hyperspectral images, but such imagery pairs with the exact
same scene are physically impossible to acquire in intraoperative settings. In
this work, we present a fully unsupervised hyperspectral image demosaicking
algorithm which only requires exemplar snapshot images for training purposes.
We regard hyperspectral demosaicking as an ill-posed linear inverse problem
which we solve using a deep neural network. We take advantage of the spectral
correlation occurring in natural scenes to design a novel inter spectral band
regularisation term based on spatial gradient consistency. By combining our
proposed term with standard regularisation techniques and exploiting a standard
data fidelity term, we obtain an unsupervised loss function for training deep
neural networks, which allows us to achieve real-time hyperspectral image
demosaicking. Quantitative results on hyperspetral image datasets show that our
unsupervised demosaicking approach can achieve similar performance to its
supervised counter-part, and significantly outperform linear demosaicking. A
qualitative user study on real snapshot hyperspectral surgical images confirms
the results from the quantitative analysis. Our results suggest that the
proposed unsupervised algorithm can achieve promising hyperspectral
demosaicking in real-time thus advancing the suitability of the modality for
intraoperative use.
| [
{
"version": "v1",
"created": "Tue, 21 Feb 2023 18:07:14 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Li",
"Peichao",
""
],
[
"Asad",
"Muhammad",
""
],
[
"Horgan",
"Conor",
""
],
[
"MacCormac",
"Oscar",
""
],
[
"Shapey",
"Jonathan",
""
],
[
"Vercauteren",
"Tom",
""
]
] | TITLE: Spatial gradient consistency for unsupervised learning of hyperspectral
demosaicking: Application to surgical imaging
ABSTRACT: Hyperspectral imaging has the potential to improve intraoperative decision
making if tissue characterisation is performed in real-time and with
high-resolution. Hyperspectral snapshot mosaic sensors offer a promising
approach due to their fast acquisition speed and compact size. However, a
demosaicking algorithm is required to fully recover the spatial and spectral
information of the snapshot images. Most state-of-the-art demosaicking
algorithms require ground-truth training data with paired snapshot and
high-resolution hyperspectral images, but such imagery pairs with the exact
same scene are physically impossible to acquire in intraoperative settings. In
this work, we present a fully unsupervised hyperspectral image demosaicking
algorithm which only requires exemplar snapshot images for training purposes.
We regard hyperspectral demosaicking as an ill-posed linear inverse problem
which we solve using a deep neural network. We take advantage of the spectral
correlation occurring in natural scenes to design a novel inter spectral band
regularisation term based on spatial gradient consistency. By combining our
proposed term with standard regularisation techniques and exploiting a standard
data fidelity term, we obtain an unsupervised loss function for training deep
neural networks, which allows us to achieve real-time hyperspectral image
demosaicking. Quantitative results on hyperspetral image datasets show that our
unsupervised demosaicking approach can achieve similar performance to its
supervised counter-part, and significantly outperform linear demosaicking. A
qualitative user study on real snapshot hyperspectral surgical images confirms
the results from the quantitative analysis. Our results suggest that the
proposed unsupervised algorithm can achieve promising hyperspectral
demosaicking in real-time thus advancing the suitability of the modality for
intraoperative use.
| no_new_dataset | 0.94887 |
2302.12002 | Sven Elflein | Sven Elflein | Master's Thesis: Out-of-distribution Detection with Energy-based Models | Master's Thesis | null | null | null | cs.LG | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Today, deep learning is increasingly applied in security-critical situations
such as autonomous driving and medical diagnosis. Despite its success, the
behavior and robustness of deep networks are not fully understood yet, posing a
significant risk. In particular, researchers recently found that neural
networks are overly confident in their predictions, even on data they have
never seen before. To tackle this issue, one can differentiate two approaches
in the literature. One accounts for uncertainty in the predictions, while the
second estimates the underlying density of the training data to decide whether
a given input is close to the training data, and thus the network is able to
perform as expected.In this thesis, we investigate the capabilities of EBMs at
the task of fitting the training data distribution to perform detection of
out-of-distribution (OOD) inputs. We find that on most datasets, EBMs do not
inherently outperform other density estimators at detecting OOD data despite
their flexibility. Thus, we additionally investigate the effects of
supervision, dimensionality reduction, and architectural modifications on the
performance of EBMs. Further, we propose Energy-Prior Network (EPN) which
enables estimation of various uncertainties within an EBM for classification,
bridging the gap between two approaches for tackling the OOD detection problem.
We identify a connection between the concentration parameters of the Dirichlet
distribution and the joint energy in an EBM. Additionally, this allows
optimization without a held-out OOD dataset, which might not be available or
costly to collect in some applications. Finally, we empirically demonstrate
that Energy-Prior Network (EPN) is able to detect OOD inputs, datasets shifts,
and adversarial examples. Theoretically, EPN offers favorable properties for
the asymptotic case when inputs are far from the training data.
| [
{
"version": "v1",
"created": "Sat, 28 Jan 2023 15:40:30 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 15:27:06 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Elflein",
"Sven",
""
]
] | TITLE: Master's Thesis: Out-of-distribution Detection with Energy-based Models
ABSTRACT: Today, deep learning is increasingly applied in security-critical situations
such as autonomous driving and medical diagnosis. Despite its success, the
behavior and robustness of deep networks are not fully understood yet, posing a
significant risk. In particular, researchers recently found that neural
networks are overly confident in their predictions, even on data they have
never seen before. To tackle this issue, one can differentiate two approaches
in the literature. One accounts for uncertainty in the predictions, while the
second estimates the underlying density of the training data to decide whether
a given input is close to the training data, and thus the network is able to
perform as expected.In this thesis, we investigate the capabilities of EBMs at
the task of fitting the training data distribution to perform detection of
out-of-distribution (OOD) inputs. We find that on most datasets, EBMs do not
inherently outperform other density estimators at detecting OOD data despite
their flexibility. Thus, we additionally investigate the effects of
supervision, dimensionality reduction, and architectural modifications on the
performance of EBMs. Further, we propose Energy-Prior Network (EPN) which
enables estimation of various uncertainties within an EBM for classification,
bridging the gap between two approaches for tackling the OOD detection problem.
We identify a connection between the concentration parameters of the Dirichlet
distribution and the joint energy in an EBM. Additionally, this allows
optimization without a held-out OOD dataset, which might not be available or
costly to collect in some applications. Finally, we empirically demonstrate
that Energy-Prior Network (EPN) is able to detect OOD inputs, datasets shifts,
and adversarial examples. Theoretically, EPN offers favorable properties for
the asymptotic case when inputs are far from the training data.
| no_new_dataset | 0.941223 |
2302.14256 | Liya Wang | Liya Wang, Alex Tien | Remote Sensing Scene Classification with Masked Image Modeling (MIM) | arXiv admin note: text overlap with arXiv:2301.12058 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | Remote sensing scene classification has been extensively studied for its
critical roles in geological survey, oil exploration, traffic management,
earthquake prediction, wildfire monitoring, and intelligence monitoring. In the
past, the Machine Learning (ML) methods for performing the task mainly used the
backbones pretrained in the manner of supervised learning (SL). As Masked Image
Modeling (MIM), a self-supervised learning (SSL) technique, has been shown as a
better way for learning visual feature representation, it presents a new
opportunity for improving ML performance on the scene classification task. This
research aims to explore the potential of MIM pretrained backbones on four
well-known classification datasets: Merced, AID, NWPU-RESISC45, and Optimal-31.
Compared to the published benchmarks, we show that the MIM pretrained Vision
Transformer (ViTs) backbones outperform other alternatives (up to 18% on top 1
accuracy) and that the MIM technique can learn better feature representation
than the supervised learning counterparts (up to 5% on top 1 accuracy).
Moreover, we show that the general-purpose MIM-pretrained ViTs can achieve
competitive performance as the specially designed yet complicated Transformer
for Remote Sensing (TRS) framework. Our experiment results also provide a
performance baseline for future studies.
| [
{
"version": "v1",
"created": "Tue, 28 Feb 2023 02:27:36 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 17:43:20 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Wang",
"Liya",
""
],
[
"Tien",
"Alex",
""
]
] | TITLE: Remote Sensing Scene Classification with Masked Image Modeling (MIM)
ABSTRACT: Remote sensing scene classification has been extensively studied for its
critical roles in geological survey, oil exploration, traffic management,
earthquake prediction, wildfire monitoring, and intelligence monitoring. In the
past, the Machine Learning (ML) methods for performing the task mainly used the
backbones pretrained in the manner of supervised learning (SL). As Masked Image
Modeling (MIM), a self-supervised learning (SSL) technique, has been shown as a
better way for learning visual feature representation, it presents a new
opportunity for improving ML performance on the scene classification task. This
research aims to explore the potential of MIM pretrained backbones on four
well-known classification datasets: Merced, AID, NWPU-RESISC45, and Optimal-31.
Compared to the published benchmarks, we show that the MIM pretrained Vision
Transformer (ViTs) backbones outperform other alternatives (up to 18% on top 1
accuracy) and that the MIM technique can learn better feature representation
than the supervised learning counterparts (up to 5% on top 1 accuracy).
Moreover, we show that the general-purpose MIM-pretrained ViTs can achieve
competitive performance as the specially designed yet complicated Transformer
for Remote Sensing (TRS) framework. Our experiment results also provide a
performance baseline for future studies.
| no_new_dataset | 0.948728 |
2303.00040 | Dezhao Luo | Dezhao Luo, Jiabo Huang, Shaogang Gong, Hailin Jin, Yang Liu | Towards Generalisable Video Moment Retrieval: Visual-Dynamic Injection
to Image-Text Pre-Training | CVPR2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The correlation between the vision and text is essential for video moment
retrieval (VMR), however, existing methods heavily rely on separate
pre-training feature extractors for visual and textual understanding. Without
sufficient temporal boundary annotations, it is non-trivial to learn universal
video-text alignments. In this work, we explore multi-modal correlations
derived from large-scale image-text data to facilitate generalisable VMR. To
address the limitations of image-text pre-training models on capturing the
video changes, we propose a generic method, referred to as Visual-Dynamic
Injection (VDI), to empower the model's understanding of video moments. Whilst
existing VMR methods are focusing on building temporal-aware video features,
being aware of the text descriptions about the temporal changes is also
critical but originally overlooked in pre-training by matching static images
with sentences. Therefore, we extract visual context and spatial dynamic
information from video frames and explicitly enforce their alignments with the
phrases describing video changes (e.g. verb). By doing so, the potentially
relevant visual and motion patterns in videos are encoded in the corresponding
text embeddings (injected) so to enable more accurate video-text alignments. We
conduct extensive experiments on two VMR benchmark datasets (Charades-STA and
ActivityNet-Captions) and achieve state-of-the-art performances. Especially,
VDI yields notable advantages when being tested on the out-of-distribution
splits where the testing samples involve novel scenes and vocabulary.
| [
{
"version": "v1",
"created": "Tue, 28 Feb 2023 19:29:05 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 17:35:07 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Luo",
"Dezhao",
""
],
[
"Huang",
"Jiabo",
""
],
[
"Gong",
"Shaogang",
""
],
[
"Jin",
"Hailin",
""
],
[
"Liu",
"Yang",
""
]
] | TITLE: Towards Generalisable Video Moment Retrieval: Visual-Dynamic Injection
to Image-Text Pre-Training
ABSTRACT: The correlation between the vision and text is essential for video moment
retrieval (VMR), however, existing methods heavily rely on separate
pre-training feature extractors for visual and textual understanding. Without
sufficient temporal boundary annotations, it is non-trivial to learn universal
video-text alignments. In this work, we explore multi-modal correlations
derived from large-scale image-text data to facilitate generalisable VMR. To
address the limitations of image-text pre-training models on capturing the
video changes, we propose a generic method, referred to as Visual-Dynamic
Injection (VDI), to empower the model's understanding of video moments. Whilst
existing VMR methods are focusing on building temporal-aware video features,
being aware of the text descriptions about the temporal changes is also
critical but originally overlooked in pre-training by matching static images
with sentences. Therefore, we extract visual context and spatial dynamic
information from video frames and explicitly enforce their alignments with the
phrases describing video changes (e.g. verb). By doing so, the potentially
relevant visual and motion patterns in videos are encoded in the corresponding
text embeddings (injected) so to enable more accurate video-text alignments. We
conduct extensive experiments on two VMR benchmark datasets (Charades-STA and
ActivityNet-Captions) and achieve state-of-the-art performances. Especially,
VDI yields notable advantages when being tested on the out-of-distribution
splits where the testing samples involve novel scenes and vocabulary.
| no_new_dataset | 0.948346 |
2303.02700 | Yujian Zheng | Yujian Zheng, Zirong Jin, Moran Li, Haibin Huang, Chongyang Ma,
Shuguang Cui, Xiaoguang Han | HairStep: Transfer Synthetic to Real Using Strand and Depth Maps for
Single-View 3D Hair Modeling | CVPR 2023 Highlight, project page:
https://paulyzheng.github.io/research/hairstep/ | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | In this work, we tackle the challenging problem of learning-based single-view
3D hair modeling. Due to the great difficulty of collecting paired real image
and 3D hair data, using synthetic data to provide prior knowledge for real
domain becomes a leading solution. This unfortunately introduces the challenge
of domain gap. Due to the inherent difficulty of realistic hair rendering,
existing methods typically use orientation maps instead of hair images as input
to bridge the gap. We firmly think an intermediate representation is essential,
but we argue that orientation map using the dominant filtering-based methods is
sensitive to uncertain noise and far from a competent representation. Thus, we
first raise this issue up and propose a novel intermediate representation,
termed as HairStep, which consists of a strand map and a depth map. It is found
that HairStep not only provides sufficient information for accurate 3D hair
modeling, but also is feasible to be inferred from real images. Specifically,
we collect a dataset of 1,250 portrait images with two types of annotations. A
learning framework is further designed to transfer real images to the strand
map and depth map. It is noted that, an extra bonus of our new dataset is the
first quantitative metric for 3D hair modeling. Our experiments show that
HairStep narrows the domain gap between synthetic and real and achieves
state-of-the-art performance on single-view 3D hair reconstruction.
| [
{
"version": "v1",
"created": "Sun, 5 Mar 2023 15:28:13 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 03:34:25 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Zheng",
"Yujian",
""
],
[
"Jin",
"Zirong",
""
],
[
"Li",
"Moran",
""
],
[
"Huang",
"Haibin",
""
],
[
"Ma",
"Chongyang",
""
],
[
"Cui",
"Shuguang",
""
],
[
"Han",
"Xiaoguang",
""
]
] | TITLE: HairStep: Transfer Synthetic to Real Using Strand and Depth Maps for
Single-View 3D Hair Modeling
ABSTRACT: In this work, we tackle the challenging problem of learning-based single-view
3D hair modeling. Due to the great difficulty of collecting paired real image
and 3D hair data, using synthetic data to provide prior knowledge for real
domain becomes a leading solution. This unfortunately introduces the challenge
of domain gap. Due to the inherent difficulty of realistic hair rendering,
existing methods typically use orientation maps instead of hair images as input
to bridge the gap. We firmly think an intermediate representation is essential,
but we argue that orientation map using the dominant filtering-based methods is
sensitive to uncertain noise and far from a competent representation. Thus, we
first raise this issue up and propose a novel intermediate representation,
termed as HairStep, which consists of a strand map and a depth map. It is found
that HairStep not only provides sufficient information for accurate 3D hair
modeling, but also is feasible to be inferred from real images. Specifically,
we collect a dataset of 1,250 portrait images with two types of annotations. A
learning framework is further designed to transfer real images to the strand
map and depth map. It is noted that, an extra bonus of our new dataset is the
first quantitative metric for 3D hair modeling. Our experiments show that
HairStep narrows the domain gap between synthetic and real and achieves
state-of-the-art performance on single-view 3D hair reconstruction.
| new_dataset | 0.954774 |
2303.03711 | Pascal Nasahl | Pascal Nasahl and Stefan Mangard | SCRAMBLE-CFI: Mitigating Fault-Induced Control-Flow Attacks on OpenTitan | Accepted at GLSVLSI'23 | null | null | null | cs.CR | http://creativecommons.org/licenses/by/4.0/ | Secure elements physically exposed to adversaries are frequently targeted by
fault attacks. These attacks can be utilized to hijack the control-flow of
software allowing the attacker to bypass security measures, extract sensitive
data, or gain full code execution. In this paper, we systematically analyze the
threat vector of fault-induced control-flow manipulations on the open-source
OpenTitan secure element. Our thorough analysis reveals that current
countermeasures of this chip either induce large area overheads or still cannot
prevent the attacker from exploiting the identified threats. In this context,
we introduce SCRAMBLE-CFI, an encryption-based control-flow integrity scheme
utilizing existing hardware features of OpenTitan. SCRAMBLE-CFI confines, with
minimal hardware overhead, the impact of fault-induced control-flow attacks by
encrypting each function with a different encryption tweak at load-time. At
runtime, code only can be successfully decrypted when the correct decryption
tweak is active. We open-source our hardware changes and release our LLVM
toolchain automatically protecting programs. Our analysis shows that
SCRAMBLE-CFI complementarily enhances security guarantees of OpenTitan with a
negligible hardware overhead of less than 3.97 % and a runtime overhead of 7.02
% for the Embench-IoT benchmarks.
| [
{
"version": "v1",
"created": "Tue, 7 Mar 2023 07:53:02 GMT"
},
{
"version": "v2",
"created": "Wed, 22 Mar 2023 12:05:50 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 10:30:09 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Nasahl",
"Pascal",
""
],
[
"Mangard",
"Stefan",
""
]
] | TITLE: SCRAMBLE-CFI: Mitigating Fault-Induced Control-Flow Attacks on OpenTitan
ABSTRACT: Secure elements physically exposed to adversaries are frequently targeted by
fault attacks. These attacks can be utilized to hijack the control-flow of
software allowing the attacker to bypass security measures, extract sensitive
data, or gain full code execution. In this paper, we systematically analyze the
threat vector of fault-induced control-flow manipulations on the open-source
OpenTitan secure element. Our thorough analysis reveals that current
countermeasures of this chip either induce large area overheads or still cannot
prevent the attacker from exploiting the identified threats. In this context,
we introduce SCRAMBLE-CFI, an encryption-based control-flow integrity scheme
utilizing existing hardware features of OpenTitan. SCRAMBLE-CFI confines, with
minimal hardware overhead, the impact of fault-induced control-flow attacks by
encrypting each function with a different encryption tweak at load-time. At
runtime, code only can be successfully decrypted when the correct decryption
tweak is active. We open-source our hardware changes and release our LLVM
toolchain automatically protecting programs. Our analysis shows that
SCRAMBLE-CFI complementarily enhances security guarantees of OpenTitan with a
negligible hardware overhead of less than 3.97 % and a runtime overhead of 7.02
% for the Embench-IoT benchmarks.
| no_new_dataset | 0.925095 |
2303.07128 | Zhou Wen | Wen Zhou | VMCDL: Vulnerability Mining Based on Cascaded Deep Learning Under Source
Control Flow | The relevant mathematical derivation has some problems such as lack
of coherence, and the location of sensitive words and the formation of slices
need to be further elaborated | null | null | null | cs.AI | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | With the rapid development of the computer industry and computer software,
the risk of software vulnerabilities being exploited has greatly increased.
However, there are still many shortcomings in the existing mining techniques
for leakage source research, such as high false alarm rate, coarse-grained
detection, and dependence on expert experience. In this paper, we mainly use
the c/c++ source code data of the SARD dataset, process the source code of
CWE476, CWE469, CWE516 and CWE570 vulnerability types, test the Joern
vulnerability scanning function of the cutting-edge tool, and propose a new
cascading deep learning model VMCDL based on source code control flow to
effectively detect vulnerabilities. First, this paper uses joern to locate and
extract sensitive functions and statements to form a sensitive statement
library of vulnerable code. Then, the CFG flow vulnerability code snippets are
generated by bidirectional breadth-first traversal, and then vectorized by
Doc2vec. Finally, the cascade deep learning model based on source code control
flow is used for classification to obtain the classification results. In the
experimental evaluation, we give the test results of Joern on specific
vulnerabilities, and give the confusion matrix and label data of the binary
classification results of the model algorithm on single vulnerability type
source code, and compare and verify the five indicators of FPR, FNR, ACC, P and
F1, respectively reaching 10.30%, 5.20%, 92.50%,85.10% and 85.40%,which shows
that it can effectively reduce the false alarm rate of static analysis.
| [
{
"version": "v1",
"created": "Mon, 13 Mar 2023 13:58:39 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 14:33:30 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Zhou",
"Wen",
""
]
] | TITLE: VMCDL: Vulnerability Mining Based on Cascaded Deep Learning Under Source
Control Flow
ABSTRACT: With the rapid development of the computer industry and computer software,
the risk of software vulnerabilities being exploited has greatly increased.
However, there are still many shortcomings in the existing mining techniques
for leakage source research, such as high false alarm rate, coarse-grained
detection, and dependence on expert experience. In this paper, we mainly use
the c/c++ source code data of the SARD dataset, process the source code of
CWE476, CWE469, CWE516 and CWE570 vulnerability types, test the Joern
vulnerability scanning function of the cutting-edge tool, and propose a new
cascading deep learning model VMCDL based on source code control flow to
effectively detect vulnerabilities. First, this paper uses joern to locate and
extract sensitive functions and statements to form a sensitive statement
library of vulnerable code. Then, the CFG flow vulnerability code snippets are
generated by bidirectional breadth-first traversal, and then vectorized by
Doc2vec. Finally, the cascade deep learning model based on source code control
flow is used for classification to obtain the classification results. In the
experimental evaluation, we give the test results of Joern on specific
vulnerabilities, and give the confusion matrix and label data of the binary
classification results of the model algorithm on single vulnerability type
source code, and compare and verify the five indicators of FPR, FNR, ACC, P and
F1, respectively reaching 10.30%, 5.20%, 92.50%,85.10% and 85.40%,which shows
that it can effectively reduce the false alarm rate of static analysis.
| no_new_dataset | 0.950134 |
2303.07768 | Dina Faneva Andriantsiory | Dina Faneva Andriantsiory, Joseph Ben Geloun, Mustapha Lebbah | DBSCAN of Multi-Slice Clustering for Third-Order Tensors | 13 pages, improved version, typos removed, text restructured, same
results | null | null | null | cs.LG stat.ML | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Several methods for triclustering three-dimensional data require the cluster
size or the number of clusters in each dimension to be specified. To address
this issue, the Multi-Slice Clustering (MSC) for 3-order tensor finds signal
slices that lie in a low dimensional subspace for a rank-one tensor dataset in
order to find a cluster based on the threshold similarity. We propose an
extension algorithm called MSC-DBSCAN to extract the different clusters of
slices that lie in the different subspaces from the data if the dataset is a
sum of r rank-one tensor (r > 1). Our algorithm uses the same input as the MSC
algorithm and can find the same solution for rank-one tensor data as MSC.
| [
{
"version": "v1",
"created": "Tue, 14 Mar 2023 10:18:31 GMT"
},
{
"version": "v2",
"created": "Sun, 19 Mar 2023 13:48:05 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 09:57:25 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Andriantsiory",
"Dina Faneva",
""
],
[
"Geloun",
"Joseph Ben",
""
],
[
"Lebbah",
"Mustapha",
""
]
] | TITLE: DBSCAN of Multi-Slice Clustering for Third-Order Tensors
ABSTRACT: Several methods for triclustering three-dimensional data require the cluster
size or the number of clusters in each dimension to be specified. To address
this issue, the Multi-Slice Clustering (MSC) for 3-order tensor finds signal
slices that lie in a low dimensional subspace for a rank-one tensor dataset in
order to find a cluster based on the threshold similarity. We propose an
extension algorithm called MSC-DBSCAN to extract the different clusters of
slices that lie in the different subspaces from the data if the dataset is a
sum of r rank-one tensor (r > 1). Our algorithm uses the same input as the MSC
algorithm and can find the same solution for rank-one tensor data as MSC.
| no_new_dataset | 0.957557 |
2303.09071 | Litao Hu | Litao Hu, Huaijin Chen, Jan P. Allebach | Joint Multi-Scale Tone Mapping and Denoising for HDR Image Enhancement | 10 pages, 4 figures, WACVW2022. Codes available at
https://github.com/hulitaotom/Joint-Multi-Scale-Tone-Mapping-and-Denoising-for-HDR-Image-Enhancement | null | 10.1109/WACVW54805.2022.00080 | null | eess.IV cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | An image processing unit (IPU), or image signal processor (ISP) for high
dynamic range (HDR) imaging usually consists of demosaicing, white balancing,
lens shading correction, color correction, denoising, and tone-mapping. Besides
noise from the imaging sensors, almost every step in the ISP introduces or
amplifies noise in different ways, and denoising operators are designed to
reduce the noise from these sources. Designed for dynamic range compressing,
tone-mapping operators in an ISP can significantly amplify the noise level,
especially for images captured in low-light conditions, making denoising very
difficult. Therefore, we propose a joint multi-scale denoising and tone-mapping
framework that is designed with both operations in mind for HDR images. Our
joint network is trained in an end-to-end format that optimizes both operators
together, to prevent the tone-mapping operator from overwhelming the denoising
operator. Our model outperforms existing HDR denoising and tone-mapping
operators both quantitatively and qualitatively on most of our benchmarking
datasets.
| [
{
"version": "v1",
"created": "Thu, 16 Mar 2023 04:04:32 GMT"
},
{
"version": "v2",
"created": "Thu, 23 Mar 2023 19:12:35 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Hu",
"Litao",
""
],
[
"Chen",
"Huaijin",
""
],
[
"Allebach",
"Jan P.",
""
]
] | TITLE: Joint Multi-Scale Tone Mapping and Denoising for HDR Image Enhancement
ABSTRACT: An image processing unit (IPU), or image signal processor (ISP) for high
dynamic range (HDR) imaging usually consists of demosaicing, white balancing,
lens shading correction, color correction, denoising, and tone-mapping. Besides
noise from the imaging sensors, almost every step in the ISP introduces or
amplifies noise in different ways, and denoising operators are designed to
reduce the noise from these sources. Designed for dynamic range compressing,
tone-mapping operators in an ISP can significantly amplify the noise level,
especially for images captured in low-light conditions, making denoising very
difficult. Therefore, we propose a joint multi-scale denoising and tone-mapping
framework that is designed with both operations in mind for HDR images. Our
joint network is trained in an end-to-end format that optimizes both operators
together, to prevent the tone-mapping operator from overwhelming the denoising
operator. Our model outperforms existing HDR denoising and tone-mapping
operators both quantitatively and qualitatively on most of our benchmarking
datasets.
| no_new_dataset | 0.951863 |
2303.09875 | Xiaotao Hu | Xiaotao Hu, Zhewei Huang, Ailin Huang, Jun Xu, Shuchang Zhou | A Dynamic Multi-Scale Voxel Flow Network for Video Prediction | CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | The performance of video prediction has been greatly boosted by advanced deep
neural networks. However, most of the current methods suffer from large model
sizes and require extra inputs, e.g., semantic/depth maps, for promising
performance. For efficiency consideration, in this paper, we propose a Dynamic
Multi-scale Voxel Flow Network (DMVFN) to achieve better video prediction
performance at lower computational costs with only RGB images, than previous
methods. The core of our DMVFN is a differentiable routing module that can
effectively perceive the motion scales of video frames. Once trained, our DMVFN
selects adaptive sub-networks for different inputs at the inference stage.
Experiments on several benchmarks demonstrate that our DMVFN is an order of
magnitude faster than Deep Voxel Flow and surpasses the state-of-the-art
iterative-based OPT on generated image quality. Our code and demo are available
at https://huxiaotaostasy.github.io/DMVFN/.
| [
{
"version": "v1",
"created": "Fri, 17 Mar 2023 10:42:05 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 03:56:20 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Hu",
"Xiaotao",
""
],
[
"Huang",
"Zhewei",
""
],
[
"Huang",
"Ailin",
""
],
[
"Xu",
"Jun",
""
],
[
"Zhou",
"Shuchang",
""
]
] | TITLE: A Dynamic Multi-Scale Voxel Flow Network for Video Prediction
ABSTRACT: The performance of video prediction has been greatly boosted by advanced deep
neural networks. However, most of the current methods suffer from large model
sizes and require extra inputs, e.g., semantic/depth maps, for promising
performance. For efficiency consideration, in this paper, we propose a Dynamic
Multi-scale Voxel Flow Network (DMVFN) to achieve better video prediction
performance at lower computational costs with only RGB images, than previous
methods. The core of our DMVFN is a differentiable routing module that can
effectively perceive the motion scales of video frames. Once trained, our DMVFN
selects adaptive sub-networks for different inputs at the inference stage.
Experiments on several benchmarks demonstrate that our DMVFN is an order of
magnitude faster than Deep Voxel Flow and surpasses the state-of-the-art
iterative-based OPT on generated image quality. Our code and demo are available
at https://huxiaotaostasy.github.io/DMVFN/.
| no_new_dataset | 0.948106 |
2303.11101 | Sungnyun Kim | Sungnyun Kim, Sangmin Bae, Se-Young Yun | Coreset Sampling from Open-Set for Fine-Grained Self-Supervised Learning | CVPR 2023 | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Deep learning in general domains has constantly been extended to
domain-specific tasks requiring the recognition of fine-grained
characteristics. However, real-world applications for fine-grained tasks suffer
from two challenges: a high reliance on expert knowledge for annotation and
necessity of a versatile model for various downstream tasks in a specific
domain (e.g., prediction of categories, bounding boxes, or pixel-wise
annotations). Fortunately, the recent self-supervised learning (SSL) is a
promising approach to pretrain a model without annotations, serving as an
effective initialization for any downstream tasks. Since SSL does not rely on
the presence of annotation, in general, it utilizes the large-scale unlabeled
dataset, referred to as an open-set. In this sense, we introduce a novel
Open-Set Self-Supervised Learning problem under the assumption that a
large-scale unlabeled open-set is available, as well as the fine-grained target
dataset, during a pretraining phase. In our problem setup, it is crucial to
consider the distribution mismatch between the open-set and target dataset.
Hence, we propose SimCore algorithm to sample a coreset, the subset of an
open-set that has a minimum distance to the target dataset in the latent space.
We demonstrate that SimCore significantly improves representation learning
performance through extensive experimental settings, including eleven
fine-grained datasets and seven open-sets in various downstream tasks.
| [
{
"version": "v1",
"created": "Mon, 20 Mar 2023 13:38:29 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 05:15:38 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Kim",
"Sungnyun",
""
],
[
"Bae",
"Sangmin",
""
],
[
"Yun",
"Se-Young",
""
]
] | TITLE: Coreset Sampling from Open-Set for Fine-Grained Self-Supervised Learning
ABSTRACT: Deep learning in general domains has constantly been extended to
domain-specific tasks requiring the recognition of fine-grained
characteristics. However, real-world applications for fine-grained tasks suffer
from two challenges: a high reliance on expert knowledge for annotation and
necessity of a versatile model for various downstream tasks in a specific
domain (e.g., prediction of categories, bounding boxes, or pixel-wise
annotations). Fortunately, the recent self-supervised learning (SSL) is a
promising approach to pretrain a model without annotations, serving as an
effective initialization for any downstream tasks. Since SSL does not rely on
the presence of annotation, in general, it utilizes the large-scale unlabeled
dataset, referred to as an open-set. In this sense, we introduce a novel
Open-Set Self-Supervised Learning problem under the assumption that a
large-scale unlabeled open-set is available, as well as the fine-grained target
dataset, during a pretraining phase. In our problem setup, it is crucial to
consider the distribution mismatch between the open-set and target dataset.
Hence, we propose SimCore algorithm to sample a coreset, the subset of an
open-set that has a minimum distance to the target dataset in the latent space.
We demonstrate that SimCore significantly improves representation learning
performance through extensive experimental settings, including eleven
fine-grained datasets and seven open-sets in various downstream tasks.
| no_new_dataset | 0.947235 |
2303.11675 | Yongkang Cheng | Yongkang Cheng, Shaoli Huang, Jifeng Ning, Ying Shan | BoPR: Body-aware Part Regressor for Human Shape and Pose Estimation | null | null | null | null | cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | This paper presents a novel approach for estimating human body shape and pose
from monocular images that effectively addresses the challenges of occlusions
and depth ambiguity. Our proposed method BoPR, the Body-aware Part Regressor,
first extracts features of both the body and part regions using an
attention-guided mechanism. We then utilize these features to encode extra
part-body dependency for per-part regression, with part features as queries and
body feature as a reference. This allows our network to infer the spatial
relationship of occluded parts with the body by leveraging visible parts and
body reference information. Our method outperforms existing state-of-the-art
methods on two benchmark datasets, and our experiments show that it
significantly surpasses existing methods in terms of depth ambiguity and
occlusion handling. These results provide strong evidence of the effectiveness
of our approach.The code and data are available for research purposes at
https://github.com/cyk990422/BoPR.
| [
{
"version": "v1",
"created": "Tue, 21 Mar 2023 08:36:59 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 08:41:24 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Cheng",
"Yongkang",
""
],
[
"Huang",
"Shaoli",
""
],
[
"Ning",
"Jifeng",
""
],
[
"Shan",
"Ying",
""
]
] | TITLE: BoPR: Body-aware Part Regressor for Human Shape and Pose Estimation
ABSTRACT: This paper presents a novel approach for estimating human body shape and pose
from monocular images that effectively addresses the challenges of occlusions
and depth ambiguity. Our proposed method BoPR, the Body-aware Part Regressor,
first extracts features of both the body and part regions using an
attention-guided mechanism. We then utilize these features to encode extra
part-body dependency for per-part regression, with part features as queries and
body feature as a reference. This allows our network to infer the spatial
relationship of occluded parts with the body by leveraging visible parts and
body reference information. Our method outperforms existing state-of-the-art
methods on two benchmark datasets, and our experiments show that it
significantly surpasses existing methods in terms of depth ambiguity and
occlusion handling. These results provide strong evidence of the effectiveness
of our approach.The code and data are available for research purposes at
https://github.com/cyk990422/BoPR.
| no_new_dataset | 0.951549 |
2303.12274 | Chunyuan Lei | Yujun Jiao, Mingze Miao, Zhishuai Yin, Chunyuan Lei, Xu Zhu, Linzhen
Nie and Bo Tao | A Hierarchical Hybrid Learning Framework for Multi-agent Trajectory
Prediction | null | null | null | null | cs.RO cs.CV | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Accurate and robust trajectory prediction of neighboring agents is critical
for autonomous vehicles traversing in complex scenes. Most methods proposed in
recent years are deep learning-based due to their strength in encoding complex
interactions. However, unplausible predictions are often generated since they
rely heavily on past observations and cannot effectively capture the transient
and contingency interactions from sparse samples. In this paper, we propose a
hierarchical hybrid framework of deep learning (DL) and reinforcement learning
(RL) for multi-agent trajectory prediction, to cope with the challenge of
predicting motions shaped by multi-scale interactions. In the DL stage, the
traffic scene is divided into multiple intermediate-scale heterogenous graphs
based on which Transformer-style GNNs are adopted to encode heterogenous
interactions at intermediate and global levels. In the RL stage, we divide the
traffic scene into local sub-scenes utilizing the key future points predicted
in the DL stage. To emulate the motion planning procedure so as to produce
trajectory predictions, a Transformer-based Proximal Policy Optimization (PPO)
incorporated with a vehicle kinematics model is devised to plan motions under
the dominant influence of microscopic interactions. A multi-objective reward is
designed to balance between agent-centric accuracy and scene-wise
compatibility. Experimental results show that our proposal matches the
state-of-the-arts on the Argoverse forecasting benchmark. It's also revealed by
the visualized results that the hierarchical learning framework captures the
multi-scale interactions and improves the feasibility and compliance of the
predicted trajectories.
| [
{
"version": "v1",
"created": "Wed, 22 Mar 2023 02:47:42 GMT"
},
{
"version": "v2",
"created": "Thu, 23 Mar 2023 09:17:51 GMT"
},
{
"version": "v3",
"created": "Fri, 24 Mar 2023 07:05:08 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Jiao",
"Yujun",
""
],
[
"Miao",
"Mingze",
""
],
[
"Yin",
"Zhishuai",
""
],
[
"Lei",
"Chunyuan",
""
],
[
"Zhu",
"Xu",
""
],
[
"Nie",
"Linzhen",
""
],
[
"Tao",
"Bo",
""
]
] | TITLE: A Hierarchical Hybrid Learning Framework for Multi-agent Trajectory
Prediction
ABSTRACT: Accurate and robust trajectory prediction of neighboring agents is critical
for autonomous vehicles traversing in complex scenes. Most methods proposed in
recent years are deep learning-based due to their strength in encoding complex
interactions. However, unplausible predictions are often generated since they
rely heavily on past observations and cannot effectively capture the transient
and contingency interactions from sparse samples. In this paper, we propose a
hierarchical hybrid framework of deep learning (DL) and reinforcement learning
(RL) for multi-agent trajectory prediction, to cope with the challenge of
predicting motions shaped by multi-scale interactions. In the DL stage, the
traffic scene is divided into multiple intermediate-scale heterogenous graphs
based on which Transformer-style GNNs are adopted to encode heterogenous
interactions at intermediate and global levels. In the RL stage, we divide the
traffic scene into local sub-scenes utilizing the key future points predicted
in the DL stage. To emulate the motion planning procedure so as to produce
trajectory predictions, a Transformer-based Proximal Policy Optimization (PPO)
incorporated with a vehicle kinematics model is devised to plan motions under
the dominant influence of microscopic interactions. A multi-objective reward is
designed to balance between agent-centric accuracy and scene-wise
compatibility. Experimental results show that our proposal matches the
state-of-the-arts on the Argoverse forecasting benchmark. It's also revealed by
the visualized results that the hierarchical learning framework captures the
multi-scale interactions and improves the feasibility and compliance of the
predicted trajectories.
| no_new_dataset | 0.947962 |
2303.12408 | Changwoon Choi | Changwoon Choi, Sang Min Kim, Young Min Kim | Balanced Spherical Grid for Egocentric View Synthesis | Accepted to CVPR 2023 | null | null | null | cs.CV | http://creativecommons.org/licenses/by/4.0/ | We present EgoNeRF, a practical solution to reconstruct large-scale
real-world environments for VR assets. Given a few seconds of casually captured
360 video, EgoNeRF can efficiently build neural radiance fields which enable
high-quality rendering from novel viewpoints. Motivated by the recent
acceleration of NeRF using feature grids, we adopt spherical coordinate instead
of conventional Cartesian coordinate. Cartesian feature grid is inefficient to
represent large-scale unbounded scenes because it has a spatially uniform
resolution, regardless of distance from viewers. The spherical parameterization
better aligns with the rays of egocentric images, and yet enables factorization
for performance enhancement. However, the na\"ive spherical grid suffers from
irregularities at two poles, and also cannot represent unbounded scenes. To
avoid singularities near poles, we combine two balanced grids, which results in
a quasi-uniform angular grid. We also partition the radial grid exponentially
and place an environment map at infinity to represent unbounded scenes.
Furthermore, with our resampling technique for grid-based methods, we can
increase the number of valid samples to train NeRF volume. We extensively
evaluate our method in our newly introduced synthetic and real-world egocentric
360 video datasets, and it consistently achieves state-of-the-art performance.
| [
{
"version": "v1",
"created": "Wed, 22 Mar 2023 09:17:01 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 08:36:20 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Choi",
"Changwoon",
""
],
[
"Kim",
"Sang Min",
""
],
[
"Kim",
"Young Min",
""
]
] | TITLE: Balanced Spherical Grid for Egocentric View Synthesis
ABSTRACT: We present EgoNeRF, a practical solution to reconstruct large-scale
real-world environments for VR assets. Given a few seconds of casually captured
360 video, EgoNeRF can efficiently build neural radiance fields which enable
high-quality rendering from novel viewpoints. Motivated by the recent
acceleration of NeRF using feature grids, we adopt spherical coordinate instead
of conventional Cartesian coordinate. Cartesian feature grid is inefficient to
represent large-scale unbounded scenes because it has a spatially uniform
resolution, regardless of distance from viewers. The spherical parameterization
better aligns with the rays of egocentric images, and yet enables factorization
for performance enhancement. However, the na\"ive spherical grid suffers from
irregularities at two poles, and also cannot represent unbounded scenes. To
avoid singularities near poles, we combine two balanced grids, which results in
a quasi-uniform angular grid. We also partition the radial grid exponentially
and place an environment map at infinity to represent unbounded scenes.
Furthermore, with our resampling technique for grid-based methods, we can
increase the number of valid samples to train NeRF volume. We extensively
evaluate our method in our newly introduced synthetic and real-world egocentric
360 video datasets, and it consistently achieves state-of-the-art performance.
| no_new_dataset | 0.952309 |
2303.12564 | Zhongjin Luo | Zhongjin Luo, Shengcai Cai, Jinguo Dong, Ruibo Ming, Liangdong Qiu,
Xiaohang Zhan, Xiaoguang Han | RaBit: Parametric Modeling of 3D Biped Cartoon Characters with a
Topological-consistent Dataset | CVPR 2023, Project page: https://gaplab.cuhk.edu.cn/projects/RaBit/ | null | null | null | cs.CV | http://creativecommons.org/publicdomain/zero/1.0/ | Assisting people in efficiently producing visually plausible 3D characters
has always been a fundamental research topic in computer vision and computer
graphics. Recent learning-based approaches have achieved unprecedented accuracy
and efficiency in the area of 3D real human digitization. However, none of the
prior works focus on modeling 3D biped cartoon characters, which are also in
great demand in gaming and filming. In this paper, we introduce 3DBiCar, the
first large-scale dataset of 3D biped cartoon characters, and RaBit, the
corresponding parametric model. Our dataset contains 1,500 topologically
consistent high-quality 3D textured models which are manually crafted by
professional artists. Built upon the data, RaBit is thus designed with a
SMPL-like linear blend shape model and a StyleGAN-based neural UV-texture
generator, simultaneously expressing the shape, pose, and texture. To
demonstrate the practicality of 3DBiCar and RaBit, various applications are
conducted, including single-view reconstruction, sketch-based modeling, and 3D
cartoon animation. For the single-view reconstruction setting, we find a
straightforward global mapping from input images to the output UV-based texture
maps tends to lose detailed appearances of some local parts (e.g., nose, ears).
Thus, a part-sensitive texture reasoner is adopted to make all important local
areas perceived. Experiments further demonstrate the effectiveness of our
method both qualitatively and quantitatively. 3DBiCar and RaBit are available
at gaplab.cuhk.edu.cn/projects/RaBit.
| [
{
"version": "v1",
"created": "Wed, 22 Mar 2023 13:46:15 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 07:49:32 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Luo",
"Zhongjin",
""
],
[
"Cai",
"Shengcai",
""
],
[
"Dong",
"Jinguo",
""
],
[
"Ming",
"Ruibo",
""
],
[
"Qiu",
"Liangdong",
""
],
[
"Zhan",
"Xiaohang",
""
],
[
"Han",
"Xiaoguang",
""
]
] | TITLE: RaBit: Parametric Modeling of 3D Biped Cartoon Characters with a
Topological-consistent Dataset
ABSTRACT: Assisting people in efficiently producing visually plausible 3D characters
has always been a fundamental research topic in computer vision and computer
graphics. Recent learning-based approaches have achieved unprecedented accuracy
and efficiency in the area of 3D real human digitization. However, none of the
prior works focus on modeling 3D biped cartoon characters, which are also in
great demand in gaming and filming. In this paper, we introduce 3DBiCar, the
first large-scale dataset of 3D biped cartoon characters, and RaBit, the
corresponding parametric model. Our dataset contains 1,500 topologically
consistent high-quality 3D textured models which are manually crafted by
professional artists. Built upon the data, RaBit is thus designed with a
SMPL-like linear blend shape model and a StyleGAN-based neural UV-texture
generator, simultaneously expressing the shape, pose, and texture. To
demonstrate the practicality of 3DBiCar and RaBit, various applications are
conducted, including single-view reconstruction, sketch-based modeling, and 3D
cartoon animation. For the single-view reconstruction setting, we find a
straightforward global mapping from input images to the output UV-based texture
maps tends to lose detailed appearances of some local parts (e.g., nose, ears).
Thus, a part-sensitive texture reasoner is adopted to make all important local
areas perceived. Experiments further demonstrate the effectiveness of our
method both qualitatively and quantitatively. 3DBiCar and RaBit are available
at gaplab.cuhk.edu.cn/projects/RaBit.
| new_dataset | 0.962778 |
2303.12930 | Tiantian Geng | Tiantian Geng, Teng Wang, Jinming Duan, Runmin Cong, Feng Zheng | Dense-Localizing Audio-Visual Events in Untrimmed Videos: A Large-Scale
Benchmark and Baseline | Accepted by CVPR2023 | null | null | null | cs.CV cs.MM cs.SD eess.AS | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ | Existing audio-visual event localization (AVE) handles manually trimmed
videos with only a single instance in each of them. However, this setting is
unrealistic as natural videos often contain numerous audio-visual events with
different categories. To better adapt to real-life applications, in this paper
we focus on the task of dense-localizing audio-visual events, which aims to
jointly localize and recognize all audio-visual events occurring in an
untrimmed video. The problem is challenging as it requires fine-grained
audio-visual scene and context understanding. To tackle this problem, we
introduce the first Untrimmed Audio-Visual (UnAV-100) dataset, which contains
10K untrimmed videos with over 30K audio-visual events. Each video has 2.8
audio-visual events on average, and the events are usually related to each
other and might co-occur as in real-life scenes. Next, we formulate the task
using a new learning-based framework, which is capable of fully integrating
audio and visual modalities to localize audio-visual events with various
lengths and capture dependencies between them in a single pass. Extensive
experiments demonstrate the effectiveness of our method as well as the
significance of multi-scale cross-modal perception and dependency modeling for
this task.
| [
{
"version": "v1",
"created": "Wed, 22 Mar 2023 22:00:17 GMT"
},
{
"version": "v2",
"created": "Fri, 24 Mar 2023 11:14:02 GMT"
}
] | 2023-03-27T00:00:00 | [
[
"Geng",
"Tiantian",
""
],
[
"Wang",
"Teng",
""
],
[
"Duan",
"Jinming",
""
],
[
"Cong",
"Runmin",
""
],
[
"Zheng",
"Feng",
""
]
] | TITLE: Dense-Localizing Audio-Visual Events in Untrimmed Videos: A Large-Scale
Benchmark and Baseline
ABSTRACT: Existing audio-visual event localization (AVE) handles manually trimmed
videos with only a single instance in each of them. However, this setting is
unrealistic as natural videos often contain numerous audio-visual events with
different categories. To better adapt to real-life applications, in this paper
we focus on the task of dense-localizing audio-visual events, which aims to
jointly localize and recognize all audio-visual events occurring in an
untrimmed video. The problem is challenging as it requires fine-grained
audio-visual scene and context understanding. To tackle this problem, we
introduce the first Untrimmed Audio-Visual (UnAV-100) dataset, which contains
10K untrimmed videos with over 30K audio-visual events. Each video has 2.8
audio-visual events on average, and the events are usually related to each
other and might co-occur as in real-life scenes. Next, we formulate the task
using a new learning-based framework, which is capable of fully integrating
audio and visual modalities to localize audio-visual events with various
lengths and capture dependencies between them in a single pass. Extensive
experiments demonstrate the effectiveness of our method as well as the
significance of multi-scale cross-modal perception and dependency modeling for
this task.
| new_dataset | 0.958304 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.