16 days ago
-
Trim dataset card
-
Update dataset card
-
Update dataset card
-
Sync README: tok_video_vjepa@224 -> tok_video_vjepa@256
-
Rename tok_video_vjepa@224 -> tok_video_vjepa@256 (VJEPA-2 features are 256px, not 224px -- fixes a naming typo)
-
Trim intro, add pretraining pointer section
-
Fix stale siglip/vjepa feature names in dataset_info metadata
-
Upload dataset
-
Note tokenized-only scope, add Visualizing the dataset section
-
Rewrite dataset card prose
-
Upload dataset
-
Upload dataset
-
Fix samples tab split (test, not train); tighten wording
-
Remove stale samples/train parquet -- preview clips are from the test split
-
Upload dataset
-
Upload dataset
-
Update sample count to 9 visual modalities; fix stale example-config link
-
Upload train/transcription
-
Upload train/tok_video_siglipv2@224
-
Upload train/human_poses
-
Upload train/tok_video_vjepa@224
-
Upload train/tok_video_opticalflow@128
-
Upload train/tok_video_normal@128
-
Upload train/det
-
Upload train/tok_video_depth@128
-
Upload train/tok_video_dinov2@224
-
Add val (test) annotation list (vol_XX/filename.mp4 label) for reconstructing raw RGB splits
-
Add train annotation list (vol_XX/filename.mp4 label) for reconstructing raw RGB splits
-
Upload test/transcription
-
Upload test/tok_video_vjepa@224
-
Upload test/tok_video_siglipv2@224
-
Upload test/tok_video_rgb@128
-
Upload test/tok_video_opticalflow@128
-
Upload test/tok_video_normal@128
-
Upload test/tok_video_dinov2@224
-
Upload test/tok_video_depth@128
-
Upload test/human_poses
-
Upload test/det
-
Upload test/crop_settings
-
Upload test/class_condition
-
Upload test/caption
-
Upload train/tok_video_rgb@128
-
Upload train/crop_settings
-
Upload train/class_condition
-
Upload train/caption
-
Update dataset card: train/test now use matching clean modality names
-
Add full dataset card (directory structure, modality mapping, samples preview)
-
Upload dataset
-
Upload dataset
-
initial commit
