Does Visual Illusion Explain Enigmatic Human Behavior?

Convolutional Neural Networks (CNNs) are a type of artificial neural network or computing system that involves the convolutional organization of neurons into receptive fields to analyze and processes visual information analogous to neurons in the visual cortex of the brain. CNNs are commonly applied in image recognition systems like face detection and recognition, autonomous vehicles, and so on.

A study on the perception of visual illusions by CNNs and human vision was conducted due to their high resemblance, by members of the Department of Information and Communication Technologies (DTIC) and the University of Valencia, published in the advanced online edition of the journal Vision Research.

Visual perception:

The visual illusion is a deceived perception of the image by the brain, that differs from the actual real image. The team trained CNNs for low-level visual tasks like denoising and deblurring. It was observed that CNNs were deceived by brightness and color illusions as similar to the human visual cortex.

This was due to the linear version and error minimization of the artificial networks. The study quotes that “these similarities are consistent with the long-standing hypothesis that considers low-level visual illusions as a by-product of the optimization to natural environments”.

The study also demonstrated that CNN’s visual perceptions were in consistent with that of humans, the reason being the enigmatic behavior of the human visual system. This was explained by Gómez Vila, first author of the study, as “for our work we also analyze when such illusions cause responses in the network that are not as physically expected, but neither do they match with human perception”.

Advertisement

The result, therefore, signified the shortcomings and disparity between the CNNs and the human visual pathway.

Source-Medindia