VidaForge is a research-oriented, end-to-end pipeline that transforms raw videos into curated, annotated, and training-ready datasets for video foundation model pretraining.

💻 Code: https://github.com/GAIR-NLP/VidaForge

📝 Blog: https://yanmaaaaaa.notion.site/vidaforge

🤗 VidaForge-3M: https://huggingface.co/datasets/VidaForge/VidaForge-3M

🎥 Introduction Video: https://www.youtube.com/watch?v=dTr7rMLEWyk