Group Member

Wei Yu
PhD. Student

Personal Homepage:  http://amp.ece.cmu.edu/people/weiyu/
Email: wy@andrew.cmu.edu

The focus of this project is to design a specialized hardware architecture for computer vision applications. Vision algorithms present great challenge for architectural design due to the high computational workload. Research on special purpose hardware has been very active in graphics and image/signal processing. There are a number of successful commercial products such as DSP, GPU. However,  no satisfying solutions for vision applications has been achieved today.

In this project, we try to find out a good interface between hardware and software based on the characters of vision tasks. To accelerate computing, dedicated hardware units for regular computing patterns will be designed. For example, specialized modules may be devoted to high-level tasks as surveillance, recognition, 3D reconstruction, or medium-level tasks such as camera calibration, motion tracking, or  low-level tasks such as filtering.  Also, parallelism will be explored on different levels, including coarse task level, fine instruction level and data level. Pipeline will be organized in a way that can favor a broad range of vision tasks. Special instruction sets and  memory hierarchy will be designed to meet the computational and bandwidth demands.

This website is maintained by Devi Parikh
Copyright © 2001 Advanced Multimedia Processing Lab. All rights reserved.
Revised: January 24, 2007