ONE-PEACE
lastmod 2023-05-25created 2023-05-243 edits
A general representation model for unlimited multimodalities
Features
- A General Represenation Model spanning vision, audio, and language modalities
- Achieves excellent results on unified tasks even without pretrained models
- Strong Emergent Zero-shot Retrieval makes it possible to align modalities not paired in the training data
- Audio-to-Image, Audtio+Text-to-Image, Audio+Image-to-Image
See also
- Deep learning
- ImageBind - Meta's multimodal AI model
- Zero-shot learning
- Few-shot learning