Deep Learning to See: Towards New Foundations of Computer Vision

The recent progress in computer vision is attributed to deep learning, but this work questions the scientific progress and proposes investigating vision within the framework of information-based laws of nature. It highlights fundamental questions about vision that remain unanswered and encourages the exploration of appropriate learning theories considering the spatiotemporal nature of the visual signal. The text is suitable for graduate and advanced undergraduate students in computer science, computational neuroscience, physics, and related disciplines.

Format: Paperback / softback
Length: 105 pages
Publication date: 27 April 2022
Publisher: Springer Nature Switzerland AG

The remarkable advancements in computer vision over the past few years have been largely attributed to the powerful combination of deep learning, fueled by the vast availability of labeled data, and the rapid growth of the GPU paradigm. While many experts subscribe to this viewpoint, this work aims to critically examine the purported scientific progress in the field and propose an alternative perspective by investigating vision within the framework of information-based laws of nature.

This work raises fundamental questions about vision that remain largely unexplored, leading the reader on a captivating journey filled with novel challenges that resonate with the foundations of machine learning. The central thesis proposed is that for a deeper understanding of visual computational processes, it is essential to look beyond the limitations of general-purpose machine learning algorithms and focus on developing appropriate learning theories that account for the spatiotemporal nature of the visual signal.

By offering a fresh perspective and challenging established beliefs, this text serves as a catalyst for inspiring and stimulating critical reflection and discussion. It is designed to be accessible to readers with no prior advanced technical knowledge, making it an ideal companion for classic textbooks on computer vision. By providing a comprehensive overview of the current state of the art, open problems, and potential solutions, this text will be of immense benefit to graduate and advanced undergraduate students in computer science, computational neuroscience, physics, and other related disciplines.

In conclusion, the remarkable progress in computer vision has been driven by the convergence of deep learning, large datasets, and GPU technology. While this perspective has contributed significantly to our understanding of vision, it is important to challenge the status quo and explore alternative viewpoints. By investigating vision within the framework of information-based laws of nature, we can uncover new insights and pave the way for innovative solutions that will shape the future of this field.

Weight: 197g
Dimension: 235 x 155 (mm)
ISBN-13: 9783030909864
Edition number: 1st ed. 2022