Hi, I'm Gabe Sarch
I'm a second-year Ph.D. student at Carnegie Mellon University in the Neural Computation program under the supervision of Dr. Mike Tarr and Dr. Katerina Fragkiadaki. My work is supported by the National Science Foundation Graduate Research Fellowship. Previously, I received a B.S. in Biomedical Engineering from the University of Rochester, where I studied the marmoset visual system under Dr. Jude Mitchell.
Self-supervision, commonsense reasoning, and active interaction in artificial agents
Animals utilize prediction and active interaction to make sense of sensory inputs without a significant amount of explicit labels or instructions. However, most state-of-the-art computer visions require millions of human annotations and are not able to generalize their previously learned knowledge to accurately reason about novel inputs or tasks. My research focuses on developing artificial agents that gain commonsense reasoning about objects and scenes by interactive and active means, drawing from psychological and neuroscientific literature when it is useful.
State-of-the-art computer vision as a model of human visual processes
State of the art computer vision systems have been shown to have high feature similarity with visual brain activity measures, such as fMRI and electrophysiology. By modeling the representations and behaviors of humans with AI systems optimized for different tasks and inputs, we can better understand how humans process naturalistic stimuli.
Publications and Preprints
Yates, J., Coop, S., Sarch, G., Wu, R., Butts, D., Rucci, M., Mitchell, J. (2021) Beyond Fixation: detailed characterization of neural selectivity in free-viewing primates bioRxiv preprint.
Fang, Z.*, Jain, A.*, Sarch, G.*, Harley, A. W., & Fragkiadaki, K. (2020). Move to See Better: Self-Improving Embodied Object Detection arXiv preprint arXiv:2012.00057. *equal contribution [project page]
See my full list of publications here