AI reconstructs visual images from human brain activity
The study utilized the open-source Stable Diffusion model, done by scientists at Osaka University in Japan. The AI model can effectively generate high-quality images, and can capture different levels of features in the image, from low-level edges and textures to high-level semantics and scenes.
They used functional magnetic resonance imaging (fMRI) to record neural activity in the human brain when viewing different types of images. Then an AI neural network is designed to learn the mapping relationship between brain activity and the potential representation of Stable Diffusion (multidimensional features of pictures). Using this network, they were able to reconstruct images from brain activity that closely resembled the original pictures.
