Visual object classification from fMRI data

Abstract

Computing devices were once limited in just calculating arithmetic. Whereas, in modern computing, complex task like object classi cation or recognition has become so popular that even our smart devices cannot be thought without having a voice, character and face recognition features. Although it has been a long time since the idea of object recognition rst came into the scene, there has been limited amount of work done in categorising objects from human fMRI data. As a result, part of human cognitive study has been neglected which possesses a large potential to be discovered and used. In brief, when a human perceives an object through vision or imagination, certain regions of brain generate speci c patterns of electric signals. Using fMRI brain data, we can potentially use those signals to interpret whatever a person is perceiving. We have tried to recreate some of the few works done previously in a limited test environment. In this paper, we try to explore an approach where a random perceived object gets split into a bunch of features it possesses. Using those extracted features, we will be able to classify the object from our previously trained deep learning model. Finally, our experiment will show a robust approach to explore and study human cognition using computers.