Making Machines That See - Geometry, Uncertainty and Deep Learning

Explore the intersection of computer vision, geometry, and deep learning in this 46-minute lecture by Professor Roberto Cipolla from the University of Cambridge. Delve into the fundamentals of 3D object reconstruction, registration, and recognition, and their commercial applications. Discover recent advancements in real-time deep learning systems that incorporate geometry and model uncertainty. Learn about Bayesian deep learning and its role in enhancing interpretability and safety in machine learning systems. Examine how geometric knowledge informs the design of effective algorithms, particularly in training networks with unlabeled data for stereo vision and human body pose and shape recovery. Gain insights into the future of computer vision, its applications in mobile technology, visual inspection, image retrieval, and video search.