Computer Vision and Pattern Recognition - EEE3032
CVPR is an undergraduate level 3 module that delivers an introduction to Computer Vision and basic Pattern Recognition and Classification topics. The module runs in Semester 2 and is on offer to all Elec. Eng. students, and to Computer Science and Maths students taking certain programmes.
It ran for the first time in 2011, receiving some of the highest MEQ student feedback scores that year (ranging 4.5-5.0) and was nominated for the Tony Jeans `Inspirational Teaching' award. The course has no relation to the eponymous vision conference.
CVPR represents an introduction to Computer Vision at undergraduate level, and M Level students might go on to take Dr. Mikolajczyk's IPV vision course. The focus of the CVPR course is to deliver a hands-on working knowledge of vision algorithms and techniques, and an understanding of the mathematics underpinning them. Consequently a series of compulsory weekly labs accompanies this module.
Topics Covered
Download the EEE3032 module specification for full details. In brief, we deliver 30 hours of lectures (3 per week) distributed over the following topics:
[1] Introduction to Computer Vision
[2-3] Image Processing
[4-7] Pattern Classification
[8-11] Features and Matching
[12-14] Shape Representation
[15-19] Tracking
[20-23] Contours Models
[24-30] Cameras and Multi-view Geometry
A powerpoint summary of the module, given on the CS options day, can be downloaded.
Sample Lectures
Most of last year's lectures were videoed as a revision aid. We have decided against public distribution due to recording quality concerns and will endeavour to film them professionally next year. Here we make of the two lectures available, to give a flavour of the material and hopefully give you an idea as to whether to select this module for the coming year.
Introduction to Splines and Curves
Introduction to Tracking
Assessment
The assessment is an 80/20 split across exam/coursework. The coursework puts into practice some of the techniques covered lectures, and this changes each year. Last year the coursework was to implement visual search, based on skeleton Matlab code.
Download the 2010/10 coursework specification for EEE3032
Background for CVPR
This module is not particularly `maths heavy' but it does require a good working knowledge of linear algebra, i.e. vectors, matrices and their various operations (dot and cross product, matrix multiplication). There is a self-test available to check if you have the mathematical background to take this module. Ignore the C programming questions as C will not be a prerequisite in this or future years.