Buckets:
Real-video dataset layout
Inside the dataset mounted at /data:
/data/
├── train.csv
└── videos/
├── 000001.mp4
├── 000002.mp4
└── ...
train.csv must contain:
path,text,num_frames,height,width,aspect_ratio,resolution,fps
Example:
path,text,num_frames,height,width,aspect_ratio,resolution,fps
000001.mp4,"A dog runs through a grassy field.",81,512,896,1.75,512,24
000002.mp4,"A car drives through a city at night.",97,576,1024,1.7778,576,24
The validation script resolves relative paths under /data/videos.
Only use videos you are authorized to use for training and redistribution.
Xet Storage Details
- Size:
- 673 Bytes
- Xet hash:
- 276372caeb7cf98cbb0ddac924be6a39f851e957c3361b23803375eb3c4ef244
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.