Seeing With Less: MindEye2’s Efficient Approach to Brain-Based Image Decoding
Patterns of fMRI activity are mapped to CLIP space and images are reconstructed with the help of our unCLIP model fine-tuned from Stable Diffusion XL.