LAION Releases the Largest Open Video Dataset for AI
Big Video Dataset (BVD) from LAION. A total of 80 million videos, 10 million hours of video, and 55 million video clips with automatically generated captions. BVD helps to improve performance on video-text tasks by 2.1 percentage points compared to InternVid.
LAION releases its Big Video Dataset (BVD). The collection of 80 million videos and 10 million hours of video, all automatically annotated with captions. The dataset is drawn from 1.3 billion CommonCrawl links. Most of the data comes from YouTube. BVD contains 55 million extracted video clips with captions and 300 million keyframes. BVD is freely available for non-commercial use and can be found on GitHub.
Models trained on BVD achieve a performance improvement of 2.1 percentage points on video-text tasks compared to InternVid. The video-text tasks involve understanding the content of videos and generating their descriptions. BVD is the biggest open video dataset to date and sets a new standard for the whole industry of video generation and video analysis.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles