Self-Calibrating Active Binocular Vision via Active Efficient Coding with Deep Autoencoders

TL;DR

This paper introduces a deep autoencoder-based model for self-calibrating active binocular vision, integrating learning of visual representations, vergence, and pursuit movements guided by an intrinsic motivation signal, demonstrating effective simulation results.

Contribution

It extends Active Efficient Coding by incorporating deep autoencoders and a novel intrinsic motivation formulation for self-calibration of binocular vision.

Findings

Successful simulation of self-calibration process

Effective learning of visual representations and eye movements

Demonstrated potential for autonomous binocular vision systems

Abstract

We present a model of the self-calibration of active binocular vision comprising the simultaneous learning of visual representations, vergence, and pursuit eye movements. The model follows the principle of Active Efficient Coding (AEC), a recent extension of the classic Efficient Coding Hypothesis to active perception. In contrast to previous AEC models, the present model uses deep autoencoders to learn sensory representations. We also propose a new formulation of the intrinsic motivation signal that guides the learning of behavior. We demonstrate the performance of the model in simulations.