Fast Geometric Libraries for Vision and Data Sciences

Explore fast geometric libraries for vision and data sciences in this 48-minute conference talk by Jean Feydy from the Finnish Center for Artificial Intelligence FCAI. Discover extensions for PyTorch, NumPy, Matlab, and R that significantly accelerate fundamental computations on generalized point clouds. Learn about breaking through computational bottlenecks in the field, with a focus on fast and scalable computations using distance matrices, efficient solvers for optimal transport problems, and applications in shape analysis and geometric deep learning. Gain insights into a case study on pixel-perfect registration of lung vessel trees. Access additional resources, including MVA lectures, videos, and papers on geometric data analysis, as well as information about the KeOps and GeomLoss libraries for geometric computations and optimal transport.