Diptychs of human and machine perceptions

TL;DR

This paper introduces diptychs combining human and machine perceptions to evaluate and compare visual understanding, highlighting key issues in current task-oriented AI.

Contribution

It presents a novel method of creating diptychs using saliency maps and human focus to qualitatively assess perception differences.

Findings

Diptychs reveal perceptual differences between humans and neural networks.

The approach offers insights into AI perception limitations.

Highlights issues in current task-oriented AI systems.

Abstract

We propose visual creations that put differences in algorithms and humans \emph{perceptions} into perspective. We exploit saliency maps of neural networks and visual focus of humans to create diptychs that are reinterpretations of an original image according to both machine and human attentions. Using those diptychs as a qualitative evaluation of perception, we discuss some crucial issues of current \textit{task-oriented} artificial intelligence.