Visual Illusions Expressed by Deep Neural Networks

Taisuke Kobayashi

doi:https://doi.org/10.36463/idw.2019.1339

1:20 PM - 1:40 PM

[PRJ6/AIS3-1(Invited)] Visual Illusions Expressed by Deep Neural Networks

*Taisuke Kobayashi¹, Eiji Watanabe^1,2 (1. Japan/Aichi/National Institute for Basic Biology (Japan), 2. Japan/Aichi/The Graduate University for Advanced Studies (SOKENDAI) (Japan))

Predictive coding, Deep neural network, Optical illusion

https://doi.org/10.36463/idw.2019.1339

The predictive coding theory, which is one of mathematical models of the visual information processing of the brain, were incorporated to deep neural networks. We found that the deep neural networks represented the motion for illusion images that were not moving physically, much like human visual perception.

Proceedings of the International Display Workshops Volume 26 (IDW '19)

[PRJ6/AIS3-1(Invited)] Visual Illusions Expressed by Deep Neural Networks