A Study and Comparison of Human and Deep Learning Recognition Performance Under Visual Distortions

TL;DR

This study compares human and deep neural network recognition performance under visual distortions, revealing significant differences and suggesting directions for developing more robust AI models.

Contribution

It provides a comparative analysis of human and DNN recognition under distortions, highlighting differences in error patterns and internal representations.

Findings

DNNs perform worse than humans on distorted images

Little correlation exists between DNN and human errors

Insights can guide the development of more robust DNNs

Abstract

Deep neural networks (DNNs) achieve excellent performance on standard classification tasks. However, under image quality distortions such as blur and noise, classification accuracy becomes poor. In this work, we compare the performance of DNNs with human subjects on distorted images. We show that, although DNNs perform better than or on par with humans on good quality images, DNN performance is still much lower than human performance on distorted images. We additionally find that there is little correlation in errors between DNNs and human subjects. This could be an indication that the internal representation of images are different between DNNs and the human visual system. These comparisons with human performance could be used to guide future development of more robust DNNs.