College of Liberal Arts
The Kersten Computational Vision Lab combines computational theory with behavioral and brain imaging experiments to understand how we see the world around us. A major theoretical challenge is to discover the computational principles required to infer the properties of objects, including their shapes, materials, and identities. This group's research uses psychophysical, brain imaging, and modeling approaches to investigate how the visual pathways of the brain transform natural image patterns into useful decisions.
The lab's current project focuses on developing a computational model for the perception of human bodies, including posture and activities that is consistent with both behavioral and neuroscience evidence. The tools include biologically based deep networks, and probabilistic modeling. However, in contrast to most models, these researchers place a strong emphasis on modeling human ability to access information about bodies across multiple levels of abstraction, including the ability to infer the 3D spatial relationships between parts over spatial scales.