Home Department: Computer Science
Supported by: Mr. Vipool M. Patel & Mrs. Sharon L. Patel
Mentor: Kalanit Grill-Spector, Psychology

Megha will seek to use computational models to explain the biological foundation of facial perception. Based on fMRI data, she hypothesizes that the size of receptive fields corresponds to our ability to distinguish human faces. By training convolutional neural networks on images of human faces, she plans to study how reducing the receptive fields in the networks affects the ability to successfully distinguish faces.

Poster presented at the Stanford Bio-X Interdisciplinary Initiatives Symposium on August 24, 2016:

Teaching a Computer to Recognize Faces: Impact of Convolutional Neural Network Architecture and Image Variations

Megha Srivastava¹, Kalanit Grill-Spector¹
[Department of Psychology¹, Stanford University]