CS6670 - Computer Vision

Picture credit: xkcd

Quick info

Instructor: Bharath Hariharan
Lecture time: Tues. and Thurs. 1:25pm - 2:40pm
Lecture venue: Phillips 219
Office Hours:
Bharath: Mon and Fri 1:25 - 2:40pm (in person at 311 Gates Hall, Zoom by appointment only )
Ed Discussions
Canvas (for zoom recordings and lecture notes)

Quick overview

This course is a graduate introduction to computer vision, and is intended to help students get started on computer vision research, or incorporate computer vision in their research.

Computer vision is the subfield of computer science that deals with the automatic analysis of visual data (i.e., images). Computer vision is historically thought of as a subset of AI, because any intelligent agent will need computer vision to perceive the world, or perceive much of the data we create. Computer vision researchers work on algorithms that take images as input and output some understanding of what is depicted, either in terms of abstract concepts (what objects are there in the image? How are they related? What people are in the image and what are they doing? etc.) or in terms of physical properties (what is the 3D shape of the depicted scene? What materials are they made of? How do they reflect light?).

Computer vision algorithms have become increasingly accurate over the past decade due to advances in machine learning techniques (specifically deep learning). This has resulted in widespread deployment of these techniques in real-world applications. This in turn has revealed several ethical quandaries, both in terms of the kinds of applications that are enabled by the technology, as well as the impact of choices in the design of these algorithms (such as training data).

This course will give students an introduction to these recent advances as well as the emerging ethical challenges. We will discuss both the latest techniques for solving specific computer vision problems, as well as attempt to situate these problems in a broader social context. Finally, this course will also focus on developing research skills, such as reading, reviewing and writing papers.

This will be a lecture-based course, with upto 2 paper readings every week. The primary deliverables will be various components of a research project. We will attempt to run a mini-computer vision conference in this course.

Intended audience and prerequisites:

This course is intended for PhD students. As such, it will assume substantial familiarity and comfort with mathematics in general and linear algebra and probability and statistics in particular. This intended audience also means that there will be a strong focus on teaching research abilities (see above for what this means).
If you are a PhD student and you find the course full, please send me a note and I will help you enroll. Undergraduate students are strongly encouraged to first take the undergraduate course (CS 4670/5670). If you have taken this course and would like to enroll, drop me a note. (I have already received many emails and we are working on the enrollment. Stay tuned).
Please look at the Syllabus below for the deliverables for this course, and the Learning Outcomes for what you should expect to learn.

Syllabus Learning outcomes Calendar