![\"\"](\"https:\/\/cdn.vanderbilt.edu\/vu-cas\/wp-content\/uploads\/sites\/58\/2024\/11\/13190533\/Frank-Tong-236x300.jpg\")
<\/p>\n<\/p>\n
Centennial Professor of Psychology
\nProfessor of Ophthalmology and Visual Sciences<\/p>\n
<\/p>\n
Date<\/strong>: Friday, November 15, 2024 Deep neural networks (DNNs) trained on tasks of object classification provide the best current models of human vision, with accompanying claims that they have attained or even surpassed human-level performance. However, DNNs tend to fail catastrophically in situations where humans do not, especially when faced with noisy, degraded, or ambiguous visual inputs. Such findings imply that the computations performed by DNNs do not adequately match those performed by the human brain. In this talk, I will evaluate the hypothesis that the robustness of both human and artificial vision depends on learning from highly challenging visual experiences. Although human vision is quite robust, just a few hours of training in the lab can promote the acquisition of more robust category-specific object representations that can generalize to novel test stimuli. Likewise, when DNNs are trained on challenge conditions of visual noise or blur, these models become more robust and human-aligned, and thereby provide better predictions of human behavioral and neural responses across diverse viewing conditions. In particular, we propose that visual blur constitutes a critical feature for conferring robustness to biological and artificial visual systems.<\/p>\n
\nTime<\/strong>:\u00a01:25PM-2:15PM
\nLocation<\/strong>:\u00a0316 Wilson Hall<\/p>\nUnderstanding the Neurocomputational Bases of Robust Object Recognition in Humans through Modeling with Deep Neural Networks<\/strong><\/h6>\n