Objectives:
The aim of this course, together with its parallel subject Deep Learning for Visual Signal Processing I, is to help the student understand and apply the theoretical and practical fundamentals behind deep learning techniques applied to image analysis and synthesis. Specifically, this course covers: i) the fundamentals and development of core image deep learning architectures, which are prevalent in almost any computer vision pipeline, ii) the basic mechanisms for generative architectures, iii) the understanding and application of visual models learning from natural language supervision and the associated image, text and coordinated representation learning techniques, and iv) an introduction to the social and environmental challenges associated deep learning computer vision; including a discussion on the implications of regulations such as the EU AI Act, addressing sustainability, interpretability, and reliability challenges of these models.