This course will review a number of computational architectures found in biological vision systems and challenging artificial vision systems, such as the computation of depth from two retinal images, the computation of motion from optic flow, mechanisms for orienting and visual attention, the analysis of complex cluttered scenes, and the recognition of complex objects. For each of these problems (and many others; see syllabus below) the major computational issues will be analyzed. A critical comparison will then be carried between biological implementations and engineering implementations derived from signal processing and computer vision.
The overall goal of this course is to provide students with an understanding of the major computational issues in vision and a critical overview of the latest advances in both computer vision and visual neuroscience. As some of the most successful computer algorithms today have direct biological inspiration, it has become essential for the engineer and scientist of tomorrow to have a broad understanding of the major computational architectures found in biological systems. This course will provide background introductory material to familiarize the student with the major challenges in fundamental neuroscience as well as in computer and biological vision. It will then critically compare the approaches employed by both fields, and survey how, in many cases, the interplay between both types of approaches has resulted in some of the most powerful artificial vision systems to date.
Applications studied in class will include the evaluation of web designs (to create web pages that are easy to use and navigate, based on the properties of the human visual system), the design of efficient advertising (either in conventional or WWW media), robot vision, the efficient compression of streaming video based on the properties of the visual system, embarked navigation aids, etc.
Beyond its topical focus on vision problems, this course will introduce students to the jargon, fundamental concepts, experimental techniques, and analysis tools used in state-of-the-art neuroscience research. This will allow students to interact with the neuroscience community, read neuroscience papers, attend neuroscience talks, and derive inspiration for new, powerful engineering algorithms.