Task 001/002 - Neuro-inspired Algorithms and Theory

ABSTRACT: Why do humans see the way that we do (with a variable resolution field of view), and how and why is this different from machines that see with uniformly spatial resolution? In the first part of the talk I will introduce the Foveation/Metamer Transform (Deza et al.; ICLR 2019) as a new image transform -- that mimics the effects of visual crowding -- was a step towards understanding if such a process known as Foveation (spatially adaptive computation) is a canonical computation required for advanced level vision akin to other known operations including convolution, half-wave rectification and pooling. In the second part of this talk, I will give an overview of similar & recent developments at the intersection of computer vision and vision science, that are thus slowly suggesting the creation of a new topic of study: Hybrid Perceptual Systems: which are systems that are part human, and part machine -- as a gateway to understand the general principles of high-level vision that go beyond the current trends in Deep Learning.

 

BIO: Arturo Deza received his B.Sc. in Robotics (Ingeniería Mecatrónica) in 2012 from Universidad Nacional de Ingeniería in Lima, Peru. He then completed his Ph.D. in Dynamical Neuroscience at the University of California, Santa Barbara in 2018 advised by Miguel Eckstein where he began his research on visual search and foveated vision in humans and machines. In 2019, Arturo moved to Harvard University as a PostDoctoral Fellow with Talia Konkle at the Department of Psychology, and from early 2020 to date he is a PostDoctoral Research Associate working with Tomaso Poggio at MIT’s Center for Brains, Minds and Machines. His research interests span across the fields of psychophysics, representation learning, human-machine perception, and recently robotics.