Biologically inspired models for computer vision applications
Reza Hojjaty Saeedy is a Ph.D. candidate working with Professor Richard A. Messner. His research field is biological signal processing in general and biologically inspired computer vision in particular. His dissertation work focuses on computational models of human visual systems and their application to computer vision and image processing. Human beings are remarkable in perceiving and analyzing natural scenes quickly and efficiently and despite recent technological development, the artificial systems are far behind the biological visual systems. Thus it is natural to look into biological vision systems as a source of inspiration for computer vision tasks. In the first part of their research, they combined several existing models of neural connectivity in the human brain to design a cascade for simulating human attention processes and detecting the salient regions in natural and synthetic scenes. Currently, they are pushing this work further to design a hybrid cascade capable of object classification. In their work, they use several machine learning methods both supervised and unsupervised.
From 2016 when Reza started his Ph.D. in IAM program, he has been a TA for courses in Math, IAM and Mechanical Engineering. In summer 2019, he spent 10 prolific weeks in Lawrence Berkeley National Lab working on probabilistic graphical models and their application in image processing. For Fall semester of 2021, he will be in an industry internship at SOLIDWORKS, MA. Outside of school, he enjoys taking road trips through New England’s natural and historic scenes.
Figure: Salient regions extracted by their algorithm (top row). A salient region in the image consists of pixels that stand out against their neighboring pixels. It is known that finding the salient points of a scene is a primary step in the process of human vision and object detection.