Designing Perceptual Puzzles by Differentiating Probabilistic Programs

TL;DR

This paper introduces a method to generate visual illusions by finding adversarial examples for probabilistic models of human perception, utilizing a differentiable probabilistic programming language to automate the creation of illusions across various visual features.

Contribution

The paper presents a novel differentiable probabilistic programming framework that enables automatic generation of perceptual illusions targeting specific visual features.

Findings

Successfully created illusions for color, size, and face perception.

Demonstrated the effectiveness of the differentiable programming approach.

Provided a new tool for studying human visual perception.

Abstract

We design new visual illusions by finding "adversarial examples" for principled models of human perception -- specifically, for probabilistic models, which treat vision as Bayesian inference. To perform this search efficiently, we design a differentiable probabilistic programming language, whose API exposes MCMC inference as a first-class differentiable function. We demonstrate our method by automatically creating illusions for three features of human vision: color constancy, size constancy, and face perception.