Representation and inference of space and motion --
dorsal stream visual processing

While there have been dramatic advances in computer vision in the areas in motion inference, tracking and structure from motion, the neural algorithms and strategeis for solving these problems are only beginning to be studied. The goal is to investigate the neural basis of hierarchical Bayesian inference along the dorsal visual pathway in the context of the problem of motion velocity tracking. We will develop a neurally realistic computational model based on Bayesian belief recurrent updating that can dynamically track the changing velocity of a sinusoidal grating and random dot patterns over time based on the activities of V1, V2 and MT neurons. We hope to gain insight into how uncertainty, prior and posterior distribution of motion velocity are being represented in the different cortical areas.

• Project 1: BBP decoding of visual motion based on responses of neurons and neuronal ensembles processes during scene analysis
  • Project Leader: Ryan Kelly
  
  
• Project 2: Recurrent computation between MT and V2
  • Project Leader: Ryan Kelly