Datasets:
Tasks:
Image Feature Extraction
Modalities:
Image
Formats:
imagefolder
Languages:
English
Size:
< 1K
ArXiv:
Tags:
climate
License:
Update README.md
Browse files
README.md
CHANGED
@@ -11,7 +11,9 @@ size_categories:
|
|
11 |
- 100M<n<1B
|
12 |
---
|
13 |
|
14 |
-
|
|
|
|
|
15 |
|
16 |
Dataset download link: https://sid.erda.dk/share_redirect/f1Hmpeh6O2
|
17 |
|
@@ -23,8 +25,10 @@ Publication: ECCV 2024 proceeding: Get Your Embedding Space in Order: Domain-Ada
|
|
23 |
|
24 |
------------------------
|
25 |
|
|
|
26 |
|
27 |
-
The DRIFT dataset includes 25k image patches collected in five European countries sourced from aerial and nanosatellite image archives.
|
|
|
28 |
|
29 |
1. Canopy height: average height value for pixels containing woody vegetation.
|
30 |
|
@@ -33,32 +37,28 @@ The DRIFT dataset includes 25k image patches collected in five European countrie
|
|
33 |
3. Tree cover fraction: percentage of the image being covered by overstory tree crowns.
|
34 |
|
35 |
|
36 |
-
The DRIFT dataset includes significant shifts between label and visual distributions due to sensor and area differences.
|
|
|
|
|
|
|
|
|
37 |
|
38 |
The dataset is a good choice for:
|
39 |
|
40 |
-
image-level regression
|
41 |
|
42 |
-
domain adaption for regression
|
43 |
|
44 |
-
remote sensing for forest applications
|
45 |
|
46 |
|
47 |
|
48 |
-
Citation:
|
49 |
|
|
|
50 |
@InProceedings{10.1007/978-3-031-72980-5_6,
|
51 |
-
author="Li, Sizhuo
|
52 |
-
and
|
53 |
-
and Brandt, Martin
|
54 |
-
and Tong, Xiaoye
|
55 |
-
and Ciais, Philippe",
|
56 |
-
editor="Leonardis, Ale{\v{s}}
|
57 |
-
and Ricci, Elisa
|
58 |
-
and Roth, Stefan
|
59 |
-
and Russakovsky, Olga
|
60 |
-
and Sattler, Torsten
|
61 |
-
and Varol, G{\"u}l",
|
62 |
title="Get Your Embedding Space in Order: Domain-Adaptive Regression for Forest Monitoring",
|
63 |
booktitle="Computer Vision -- ECCV 2024",
|
64 |
year="2024",
|
|
|
11 |
- 100M<n<1B
|
12 |
---
|
13 |
|
14 |
+
|
15 |
+
|
16 |
+
# The DRIFT (Domain-Adaptive Regression for Forest Monitoring) dataset
|
17 |
|
18 |
Dataset download link: https://sid.erda.dk/share_redirect/f1Hmpeh6O2
|
19 |
|
|
|
25 |
|
26 |
------------------------
|
27 |
|
28 |
+
## Description
|
29 |
|
30 |
+
The DRIFT dataset includes 25k image patches collected in five European countries sourced from aerial and nanosatellite image archives.
|
31 |
+
Each image patch is associated with three target variables to predict:
|
32 |
|
33 |
1. Canopy height: average height value for pixels containing woody vegetation.
|
34 |
|
|
|
37 |
3. Tree cover fraction: percentage of the image being covered by overstory tree crowns.
|
38 |
|
39 |
|
40 |
+
The DRIFT dataset includes significant shifts between label and visual distributions due to sensor and area differences.
|
41 |
+
Furthermore, vegetation tends to grow to fit the local climate, therefore introducing concept drift in the data: same tree species may appear differently in different subsets. The label distribution also varies among different subsets (countries).
|
42 |
+
|
43 |
+
|
44 |
+
## Applications
|
45 |
|
46 |
The dataset is a good choice for:
|
47 |
|
48 |
+
* image-level regression
|
49 |
|
50 |
+
* domain adaption for regression
|
51 |
|
52 |
+
* remote sensing for forest applications
|
53 |
|
54 |
|
55 |
|
56 |
+
## Citation:
|
57 |
|
58 |
+
```
|
59 |
@InProceedings{10.1007/978-3-031-72980-5_6,
|
60 |
+
author="Li, Sizhuo and Gominski, Dimitri and Brandt, Martin and Tong, Xiaoye and Ciais, Philippe",
|
61 |
+
editor="Leonardis, Ale{\v{s}} and Ricci, Elisa and Roth, Stefan and Russakovsky, Olga and Sattler, Torsten and Varol, G{\"u}l",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
title="Get Your Embedding Space in Order: Domain-Adaptive Regression for Forest Monitoring",
|
63 |
booktitle="Computer Vision -- ECCV 2024",
|
64 |
year="2024",
|