GPU-based Computer Vision and Pattern Recognition Library

Research areas

Temporary Supervisor

Dr Jun Zhou

Description

In this project, students are required to develop a library in computer vision and pattern recognition using CUDA C/C++ for GPU computing. The library will run on NVIDIA Tesla GPU Unit, which is a high performance parallel computing system with multiple parallel processing cores. Students can pick from a variety of computer vision and pattern recognition methods for implementation. It is expected that the parallel computing implementation can accelerate these methods by an order of magnitude or more. If time permitting, students can apply their algorithms/codes to real-world applications. The library will be divided into smaller modules, thus, several students can work simultaneously on the project.

Goals

Develop modules for a GPU-based Computer Vision and Pattern Recognition Library.

Requirements

C/C++ programming experiences. Basic knowledge on computer vision and pattern recognition, or on CUDA computing.

Background Literature

Linda G. Shapiro and George C. Stockman, "Computer Vision", Prentice Hall, 2001. Jason Sanders and Edward Kandrot, "CUDA by Example: An Introduction to General-Purpose GPU Programming", Addison-Wesley, 2010.

Gain

Students can get familiar with programming under NVIDIAs CUDA architecture, as well as fundamental methods in computer vision and pattern recognition.

Updated:  1 June 2019/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing