Abstract
The development of binocular vision is an active learning process comprising the development of disparity tuned neurons in visual cortex and the establishment of precise vergence control of the eyes. We present a computational model for the learning and self-calibration of active binocular vision based on the Active Efficient Coding framework, an extension of classic efficient coding ideas to active perception. Under normal rearing conditions with naturalistic input, the model develops disparity tuned neurons and precise vergence control, allowing it to correctly interpret random dot stereograms. Under altered rearing conditions modeled after neurophysiological experiments, the model qualitatively reproduces key experimental findings on changes in binocularity and disparity tuning. Furthermore, the model makes testable predictions regarding how altered rearing conditions impede the learning of precise vergence control. Finally, the model predicts a surprising new effect that impaired vergence control affects the statistics of orientation tuning in visual cortical neurons.
Highlights
Humans and other species learn to perceive the world largely autonomously
In contrast, we aim to demonstrate the generality of the Active Efficient Coding (AEC) approach by reproducing and explaining a large range of neurophysiological findings from different alternate rearing conditions: changing the orientation distribution in the visual input, monocular rearing, strabismic rearing, and aniseikonia
Our results show that the model qualitatively captures findings on how different alternate rearing conditions alter the statistics of disparity tuning and binocularity
Summary
Humans and other species learn to perceive the world largely autonomously. This is in sharp contrast to today’s machine learning approaches (Kotsiantis et al, 2007; Jordan and Mitchell, 2015), which typically use millions of carefully labeled training images in order to learn to, say, recognize an object or perceive its three-dimensional structure. Species with two forward facing eyes learn to register small differences between the images projected onto the left and right retinas These differences are called binocular disparities and are detected by populations of neurons in visual cortex (Kandel et al, 2000; Blake and Wilson, 2011) that have receptive subfields in both eyes. They are modeled using separate Gabor-shaped filters for each eye, where the disparity is encoded by a shift in the centers of the filters, a difference between their phases, or by a combination of both (Fleet et al, 1996; Chen and Qian, 2004). This learning does not require any supervision from outside, but must rely on some form of selfcalibration
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call