What AI says about ImageBind

ImageBind is a PyTorch-based project from Facebook AI Research that provides implementations and pretrained models for learning a single joint embedding space that binds six modalities: images, text, audio, depth, thermal, and IMU data. The unified embeddings enable out-of-the-box cross-modal capabilities, including cross-modal retrieval, composing modalities with arithmetic, and cross-modal detection and generation. The work is described in the paper 'ImageBind: One Embedding Space To Bind Them All' and is accompanied by demos, model cards, and usage guidance.

Hosted on GitHub