Spaces:
Running
on
Zero
Running
on
Zero
File size: 2,705 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 |
# Copyright (c) 2025 Ye Liu. Licensed under the BSD-3-Clause License.
import nncore
from videomind.dataset.hybrid import DATASETS
from videomind.dataset.wrappers import AnsweringCropDataset, AnsweringDataset, GroundingDataset
from videomind.utils.parser import parse_query, parse_question
@DATASETS.register(name='rextime')
class ReXTimeDataset(AnsweringDataset):
ANNO_PATH_TRAIN = 'data/rextime/rextime_train.json'
ANNO_PATH_VALID = 'data/rextime/rextime_val.json'
ANNO_PATH_TEST = 'data/rextime/rextime_test_release.json'
VIDEO_ROOT_ANET = 'data/activitynet/videos_3fps_480_noaudio'
VIDEO_ROOT_QVHL = 'data/qvhighlights/videos_3fps_480_noaudio'
DURATIONS_ANET = 'data/activitynet/durations.json'
DURATIONS_QVHL = 'data/qvhighlights/durations.json'
SOURCE = 'rextime'
DATA_TYPE = 'multimodal'
UNIT = 1.0
MIN_LEN = 64
@classmethod
def load_annos(self, split='train'):
if split == 'train':
raw_annos = nncore.load(self.ANNO_PATH_TRAIN)
elif split == 'valid':
raw_annos = nncore.load(self.ANNO_PATH_VALID)
else:
print('WARNING: Test split does not have ground truth annotations')
raw_annos = nncore.load(self.ANNO_PATH_TEST)
durations_anet = nncore.load(self.DURATIONS_ANET)
durations_qvhl = nncore.load(self.DURATIONS_QVHL)
annos = []
for raw_anno in raw_annos:
vid = raw_anno['vid']
if len(vid) == 13:
video_path = nncore.join(self.VIDEO_ROOT_ANET, vid + '.mp4')
duration = durations_anet[vid]
else:
video_path = nncore.join(self.VIDEO_ROOT_QVHL, vid + '.mp4')
duration = durations_qvhl[vid]
anno = dict(
source=self.SOURCE,
data_type=self.DATA_TYPE,
video_path=video_path,
duration=duration,
query=parse_query(raw_anno['question']),
question=parse_question(raw_anno['question']),
options=[o.capitalize() for o in raw_anno['options']],
answer=raw_anno['answer'].replace('From <s0> to <e0>, ', '').capitalize(),
ans=raw_anno['ans'],
span=[raw_anno['span']],
task=raw_anno['category'])
annos.append(anno)
return annos
@DATASETS.register(name='rextime_crop')
class ReXTimeCropDataset(AnsweringCropDataset, ReXTimeDataset):
SOURCE = 'rextime_crop'
@DATASETS.register(name='rextime_grounding')
class ReXTimeGroundingDataset(GroundingDataset, ReXTimeDataset):
SOURCE = 'rextime_grounding'
DATA_TYPE = 'grounding'
|