Laion Big Video Dataset
The article introduces LAION-BVD, a massive open-source video dataset containing 10 million hours of video for multimodal machine learning research. The researchers demonstrate that models trained on this data achieve competitive performance across video, audio, and image-text benchmarks.
Why it matters
Large-scale open datasets are essential for democratizing AI research and advancing the capabilities of multimodal generative models.
A 10-Million-Hour Open Web Video Corpus for Video, Audio, and Image-Text Pre-training
Read Paper Download Dataset GitHub * Shared first authors ◇ Shared last authors 1 Tübingen AI Center, University of Tübingen 2 LAION 3 JSC, FZJ 4 Wynd Labs 5 MPI for Intelligent Systems, ELLIS Institute Tübingen 6 MCML, Technical University Munich Affiliated Institutions
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in