Modern Computer Vision

Computer vision today is driven by deep networks, but the way we design, train, and reason about them is deeply rooted in classical vision and signal processing. To truly do research and build creative new methods, you need both: learning and understanding pixels. MCV is roughly 70% deep learning for vision and 30% the classical foundations that enable it.

We go from the very basic math all the way to state-of-the-art practice. The goal is to get to the bottom of things, develop real intuition, and encourage creative new thinking. Five homework sets each have three tracks: theory, from-scratch implementation, and applied. The course has a research orientation and concludes with a final research project.