A lot of computation takes place from the image falling on our retina to the brain forming a cognitive map. An important computation is inferring the geometry structure of scenes. Here we let people watch videos of cars navigating in virtual towns, and modeled the synchronized neural activity across people while varying the weather and lighting conditions in the videos. We found that two types of representation: a compact representation of 3D geometry structure, and the representation of road types, are encoded in a wide range of brain regions spanning three pathways, while the representation of road types is absent in early visual cortex. Check out the preprint!
Cai Lab at University of Miami