By Chetan Ralekar, Massachusetts Institute of Technology (MIT) Cambridge, MA 02139, USA

Abstract 

In computer vision, pattern recognition is often treated as a machine optimization problem, with limited use of insights from human cognition. In this talk, I will discuss cognitively inspired approaches to pattern recognition, focusing on character and face recognition. I will present how behavioral and eye-tracking studies revealed the information-pickup strategies humans use during character recognition, providing insights for designing more robust deep learning-based recognition models. By integrating human eye-movement data with explainable AI techniques, we developed training strategies that significantly improved recognition, particularly for degraded inputs, helping bridge the gap between human and machine learning. I will also briefly discuss our work on the neural mechanisms of visual learning in children who experienced early visual deprivation, highlighting how visual experience shapes brain development. Finally, I will conclude by outlining broader research directions at the intersection of cognitive neuroscience, vision science, and artificial intelligence.