Spaces:
Running
on
Zero
Running
on
Zero
File size: 3,416 Bytes
6073e55 23fdbc0 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 |
# Copyright (c) 2025 Ye Liu. Licensed under the BSD-3-Clause License.
from collections import OrderedDict
import nncore
from videomind.dataset.hybrid import DATASETS
from videomind.dataset.wrappers import GroundingDataset
from videomind.utils.parser import parse_query
@DATASETS.register(name='youcook2')
class YouCook2Dataset(GroundingDataset):
ANNO_PATH = 'data/youcook2/youcookii_annotations_trainval.json'
VIDEO_ROOT = 'data/youcook2/videos_3fps_480_noaudio'
UNIT = 1.0
@classmethod
def load_annos(self, split='train'):
subset = 'training' if split == 'train' else 'validation'
raw_annos = nncore.load(self.ANNO_PATH, object_pairs_hook=OrderedDict)['database']
all_videos = nncore.ls(self.VIDEO_ROOT, ext='.mp4')
all_videos = set(v[:11] for v in all_videos)
annos = []
for vid, raw_anno in raw_annos.items():
if raw_anno['subset'] != subset:
continue
if vid not in all_videos:
continue
duration = raw_anno['duration']
for meta in raw_anno['annotations']:
anno = dict(
source='youcook2',
data_type='grounding',
video_path=nncore.join(self.VIDEO_ROOT, vid + '.mp4'),
duration=duration,
query=parse_query(meta['sentence']),
span=[meta['segment']])
annos.append(anno)
annos.append(anno)
return annos
@DATASETS.register(name='youcook2_bias')
class YouCook2BiasDataset(YouCook2Dataset):
@classmethod
def load_annos(self, split='train'):
subset = 'training' if split == 'train' else 'validation'
raw_annos = nncore.load(self.ANNO_PATH, object_pairs_hook=OrderedDict)['database']
all_videos = nncore.ls(self.VIDEO_ROOT, ext='.mp4')
all_videos = set(v[:11] for v in all_videos)
annos = []
for vid, raw_anno in raw_annos.items():
if raw_anno['subset'] != subset:
continue
if vid not in all_videos:
continue
duration = raw_anno['duration']
moments = raw_anno['annotations']
for i in range(len(moments) - 1):
span_a = moments[i]['segment']
span_b = moments[i + 1]['segment']
if span_b[0] - span_a[1] < 3:
query_a = parse_query(f"The moment before {moments[i + 1]['sentence']}")
query_b = parse_query(f"The moment after {moments[i]['sentence']}")
anno_a = dict(
source='youcook2_bias',
data_type='grounding',
video_path=nncore.join(self.VIDEO_ROOT, vid + '.mp4'),
duration=duration,
query=parse_query(query_a),
span=[span_a])
anno_b = dict(
source='youcook2_bias',
data_type='grounding',
video_path=nncore.join(self.VIDEO_ROOT, vid + '.mp4'),
duration=duration,
query=parse_query(query_b),
span=[span_b])
annos.append(anno_a)
annos.append(anno_b)
return annos
|