Contrast Sensitivity Functions in Autoencoders

TL;DR

This paper investigates how convolutional autoencoders trained on low-level vision tasks can develop human-like contrast sensitivity functions, highlighting the limitations and implications of using CNNs for modeling human visual perception.

Contribution

It demonstrates that CNN autoencoders can replicate human CSFs for certain tasks and shows that deeper networks may perform worse in modeling low-level visual phenomena.

Findings

Autoencoders can develop human-like CSFs in specific low-level tasks.

Deeper CNNs may be less effective in replicating human low-level vision phenomena.

CNN architectures' simplicity can limit their usefulness in understanding human vision.

Abstract

Three decades ago, Atick et al. suggested that human frequency sensitivity may emerge from the enhancement required for a more efficient analysis of retinal images. Here we reassess the relevance of low-level vision tasks in the explanation of the Contrast Sensitivity Functions (CSFs) in light of (1) the current trend of using artificial neural networks for studying vision, and (2) the current knowledge of retinal image representations. As a first contribution, we show that a very popular type of convolutional neural networks (CNNs), called autoencoders, may develop human-like CSFs in the spatio-temporal and chromatic dimensions when trained to perform some basic low-level vision tasks (like retinal noise and optical blur removal), but not others (like chromatic adaptation or pure reconstruction after simple bottlenecks). As an illustrative example, the best CNN (in the considered set of simple architectures for enhancement of the retinal signal) reproduces the CSFs with an RMSE error of 11\% of the maximum sensitivity. As a second contribution, we provide experimental evidence of the fact that, for some functional goals (at low abstraction level), deeper CNNs that are better in reaching the quantitative goal are actually worse in replicating human-like phenomena (such as the CSFs). This low-level result (for the explored networks) is not necessarily in contradiction with other works that report advantages of deeper nets in modeling higher-level vision goals. However, in line with a growing body of literature, our results suggests another word of caution about CNNs in vision science since the use of simplified units or unrealistic architectures in goal optimization may be a limitation for the modeling and understanding of human vision.