LAION has released a substantial open video dataset aimed at advancing AI research, featuring 80 million videos and over 10 million hours of footage, according to The Decoder. This dataset also includes 55 million clips with automatic descriptions, providing a rich resource for training AI models.
Models trained using LAION’s Big Video Dataset (BVD) have demonstrated improved performance, surpassing the previous benchmark set by InternVid by up to 2.1 percentage points, The Decoder reported. This marks a notable step forward in video-based AI capabilities.
The release is supported by a 2024 ruling from a Hamburg court, which permits the collection of copyrighted content for non-commercial research purposes. For Japanese markets, this development signals expanding opportunities in AI-driven video analysis, relevant to sectors like finance and media where video data is increasingly leveraged.