Designing Perceptual Puzzles by Differentiating Probabilistic Programs
This work addresses the challenge of generating perceptual puzzles for researchers in cognitive science and vision, though it appears incremental as it applies a novel method to an existing domain.
The authors tackled the problem of designing visual illusions by finding adversarial examples for Bayesian models of human perception, resulting in an automated method that created illusions for color constancy, size constancy, and face perception.
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.