Private dataset. Mixed-provenance research data. Not cleared for public
redistribution — see Provenance & License before changing visibility.
Generated by scripts/build_video_training_dataset.py.
Videos: 648 (≈0.94 h, ~1.9 GiB)
NPZ training pairs: 79 (causal_forcing: 33, pframe_npz: 46; ~12 GiB)
Total payload: ~13.9 GiB
Splits — train: 577, val: 36, test: 35
This dataset mixes several… See the full description on the dataset page:
https://huggingface.co/datasets/simbahuang/video_training_v1.
