Obsidian/Paper/Diffusion/MoMA.md

方法：

1. Multimodal Generative Image-feature Decoder
2. Self-Attention Feature Transfer
3. Multimodal Generative Learning and Diffusion Learning