Human Body Capture: Motion Reconstruction and Synthesis

This lecture covers topics related to human body capture, motion reconstruction, and synthesis in 3D scenes. It discusses markerless human motion capture, learning motion priors from mocap datasets, and reconstructing natural human motions with monocular cameras. The lecture also explores the creation of 4D human body capture models, egocentric human behavior capture, and human body shape and motion analysis from head-mounted devices. Additionally, it delves into compositional human-scene interaction synthesis with semantic control, generative human motion models, and 3D segmentation of humans in point clouds using synthetic data.