Deep Learning to See: Towards New Foundations of Computer Vision

TL;DR

This paper critiques the current scientific understanding of computer vision's progress, advocating for a shift towards information-based laws and specialized learning theories that consider the spatiotemporal nature of visual signals.

Contribution

It proposes a new foundational approach to computer vision grounded in information theory and emphasizes the need for specialized learning frameworks beyond general machine learning algorithms.

Findings

Challenges the sufficiency of current deep learning approaches

Highlights the importance of spatiotemporal information in visual processing

Calls for new theoretical foundations in computer vision

Abstract

The remarkable progress in computer vision over the last few years is, by and large, attributed to deep learning, fueled by the availability of huge sets of labeled data, and paired with the explosive growth of the GPU paradigm. While subscribing to this view, this book criticizes the supposed scientific progress in the field and proposes the investigation of vision within the framework of information-based laws of nature. Specifically, the present work poses fundamental questions about vision that remain far from understood, leading the reader on a journey populated by novel challenges resonating with the foundations of machine learning. The central thesis is that for a deeper understanding of visual computational processes, it is necessary to look beyond the applications of general purpose machine learning algorithms and focus instead on appropriate learning theories that take into account the spatiotemporal nature of the visual signal.