Advances in Computer Vision
MIT OpenCourseWare offers this advanced computer vision course covering three major areas. The geometry unit spans image formation, representation theory, classic and deep-learning-based multi-view geometry, differentiable rendering, neural scene representations, correspondence estimation, optical flow, and point tracking. A second unit covers generative modeling and representation learning, including image and video generation, diffusion model guidance, conditional probabilistic models, and contrastive and masking-based representation learning. The final unit examines vision for embodied agents, looking at how visual perception feeds into decision-making, planning, and control in robotics. Materials follow MIT's OCW format of lecture notes and course structure released under a Creative Commons license, free to access with no certificate offered. The course targets students who already have grounding in computer vision fundamentals and want the current research frontier.