Low-Level Visual Psychophysics: A Bridge Between Psychologist and Deep Learning Engineers

1 June 2023, 1.00 PM - 1 June 2023, 2.00 PM

Jesús Malo (Associate Professor, School of Physics, University of Valencia, Spain)

Room 2D17, staff common room, School of Psychological Science, Priory Road and online

Hosted by the Generalisation in Mind & Machine research group

Full details can be found on the Mind and Machine website: https://mindandmachine.blogs.bristol.ac.uk/seminars/

Abstract: Conventional training of artificial neural networks on benchmarks is not a good strategy for developing models of human vision. Specifically, Bowers et al. BBS 2022 review many failures of such a strategy both in physiology (neural recordings) and in high-level behaviour (Gestalt, classification, and more). As a result, the reader may end up with an overly pessimistic view of the prospects for deep learning in this field. This can steer these modelers away from human vision problems. The separation of the Machine Learning and Vision Science communities would take eventually valuable input away from visual neuroscience,  and would also be a problem for machine vision because it could benefit from psychology. 

In this talk I will argue that low-level visual psychophysics (e.g. colour and texture discrimination, image quality…) can be a bridge between both communities. On the one hand, I will show that pattern discrimination is good enough to point out the limitations of the naive benchmark-only approach [1]: it suggests specific changes in architecture which are not obvious from conventional practices in deep learning [1,2], where deeper deeper is not more human [3-5]. However, on the other hand, the low dimensionality of these low-level (image-patch) perception problems is an advantage too: I will show that the behaviour of artificial networks and image-probability models for patches is easier to compare with (and to benefit from!) classical early vision models [6]. In this way, low-level visual psychophysics can serve as a common ground for positive interaction between the Machine Learning and the Vision Science communities: it represents an independent way to evaluate artificial systems trained for some other tasks, classical vision models can inspire modifications in the architectures, and non-Euclidean perceptual metrics can serve as effective regularisers (or priors) when there is not enough data to train machine learning models. 

Short Biography:  Jesús Malo received the M.Sc. degree in Physics in 1995 and the Ph.D. degree in Physics in 1999 both from the Universitat de València.  He was the recipient of the Vistakon European Research Award in 1994. In 2000 and 2001 he worked as Fulbright Postdoc at the Vision Group of the NASA Ames Research Center (A.B. Watson), and at the Lab of Computational Vision of the Center for Neural Science, New York University (E.P. Simoncelli). He came back to the NYU in 2013 for a semester. Currently, he serves as Associate Editor of IEEE Trans. Im. Proc. He is with the Image and Signal Processing Group and the Visual Statistics Group (VI(S)TA) at the Universitat de València. He is member of the Asociación de Mujeres Investigadoras y Tecnólogas (AMIT).

If joining online: https://bristol-ac-uk.zoom.us/j/93295465286?pwd=MTVhZ2NMcVFiaG5DeUlybTRLVjcxZz09, Meeting ID: 932 9546 5286 | Passcode: 074795

Contact information

Contact Abla Hatherell with any enquiries. 

Edit this page