DREAM: Visual Decoding from REversing HumAn Visual SysteM (Paper Review)

Date:

A presentation on an fmri-to-image method that generates reconstructed images using functional MRI data. Patient on MRI scanner is shown images and BOLD activity in visual cortex is collected. Deep learning models are trained to encode semantics, color profile, and depth maps to be fed to a pretrained diffusion model. Press image below for full PDF. DL4D1