Computer Vision

CS 554 treats vision as an inference problem: given pixels, recover the geometry, motion, and semantic structure of the scene that produced them, drawing on both the math (linear algebra, probability, optimization) and a working sense of how human perception solves the same task. You'll work through a homework set on classical pipelines (features, camera geometry, stereo), survey a slice of the recent literature, and carry a project from problem formulation to experimental analysis, typically leaning on deep learning by the second half. It sits on top of a solid linear algebra and ML background and feeds directly into research or applied work in robotics, graphics, medical imaging, and anything else where the input is an image or a video stream.