MENTAL IMAGE TO REAL APPEARANCE VIA NEURAL NETWORK

Views: 374
BitClave BitClave BitClave

Researchers from the University of Kyoto reported about the results of the deep reconstruction of images based on analysis of brain activity. With the help of functional magnetic resonance imaging, they created a decoding method that allows converting the visual activity of the brain into the hierarchical functions of a deep neural network. It turned out that the brain processes the visual information hierarchically, highlighting different levels and components with different levels of complexity. The method of reconstruction of thoughtforms allowed to see both real objects that a person sees at a given moment, and objects from memory. The results of the research show that the hierarchical visual information of the brain can be effectively combined to reconstruct perceived and real images.

Comments: 0