PredNet Neural Network Reproduces Human Visual Illusions

Eiji Watanabe and colleagues recently found that PredNet, a deep predictive neural network, reproduces the rotating snakes visual illusion, interpreting static concentric patterns as motion. Published in Frontiers in Psychology, the study supports predictive coding models of perception and suggests deep networks can both probe perceptual mechanisms and generate new illusions, offering a computational tool for neuroscience and vision science research.