The World's Largest Decentralized AGI Multimodal Dataset
GitHub: https://github.com/omegalabsinc/omegalabs-bittensor-subnetMiner:
- Performs a simple search on YouTube and retrieves 8 videos at a time
- Provides a certain clip range (maximum of 2 minutes) and a description (catch) which includes the title, tags, and description of the video.
- Obtains the ImageBind embeddings for the video, audio, and caption.
- Returns the video ID, caption, ImageBind embeddings (video, audio, caption embeddings), and start and end times for the clips (maximum of 2 minutes).