The National Eye Institute (NEI) awarded a five-year, $409,375 Project Grant (CFDA 93.867 - Vision Research) to The Johns Hopkins University to decipher brain algorithms underlying 3D object perception and develop artificial vision networks to achieve human-level 3D visual performance.
The research aims to use artificial neural networks to replicate the 3D shape tuning functions of individual neurons recorded in the visual cortex of macaque monkeys. The researchers will analyze the 3D shape processing algorithms learned by these neuron-trained networks, tracing the information pathways from early visual processing to high-level object recognition. They will then incorporate the deciphered biological algorithms into novel artificial network architectures designed for internal 3D shape reconstruction and inference in natural scenes. This work leverages a photorealistic 3D shape stimulus dataset produced by genetic algorithms to support training and testing across larger 3D shape domains under real-world viewing conditions.