Computer Vision

Computer vision (CV) is an exciting and rapidly changing field. In a little over a dozen years, deep learning algorithms have revolutionized many aspects of CV. In the past few years, the emergence of powerful generative algorithms are having a significant impact on CV. Applications that were infeasible or impractical a few years ago are now in routine production. These advances allow intelligent systems to interact with the real-world using CV. Examples of modern CV applications include digital photography, robotic vision, autonomous vehicles, image generation, medical imaging, and scientific imaging. This course is a fast-paced survey of both the fundamental theory of CV algorithms and hands-on practice. Successfully deploying CV applications often requires a combination of classical methods and state-of-the-art algorithms. Therefore, this course covers the classical image processing and CV techniques, often found in practical CV pipelines. From this foundation the course moves to the predictive and generative deep learning algorithms, that have revolutionized CV. Students apply tools drawn from the extensive universe of Python CV-related packages in the hands-on assignments to reinforce key principles. Students enrolled for graduate credit are required to perform, present, and report on an independent project. This project must demonstrate a mastery of methods covered in the course as applied to a suitable real-world data set. Major topics covered in the course include algorithms used to prepare and transform images and extract features; statistical properties of images and methods of decomposition; components of deep learning algorithms for CV; classification of objects in images; generative CV algorithms; motion in images and optical flow; object detection and tracking algorithms; models for 3D vision; and segmentation of images.