 |
About AMP Lab Projects Downloads Publications People Links
Group Member
 |
Dhiraj Goel Personal Homepage:
http://amp.ece.cmu.edu/people/Dhiraj/ |
|
|
Office: Hamerschlag Hall D-Level C8 Lab: Porter Hall B6 Phone: 412-268-4264 Fax: 412-268-3890 |
Mailing Address: Department of ECE, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213-3890 |
|
[Research Interests] [Project] [Publications]
Interest Areas:
- Computer Vision
- Pattern Recognition/Machine Learning
- Embedded Vision Systems
- Computer Graphics
- Image Processing & Multimedia
Current Research Focus:
-
Algorithms for detecting pedestrians in a video using motion patterns
We aim to build robust and real-time pedestrian detection systems by analyzing human motion patterns. This approach has many advantages over detection by shape alone, esp. in cluttered environments.
-
CMUcam3: Low cost embedded vision system
CMUcam3 is a cheap, low powered embedded system providing vision capabilities in an easy to use environment.
Top of this page
|
Algorithms for detecting pedestrians in a video using motion patterns
Due to high intra-class variability, pedestrian detection using shape alone is a very hard problem. In this project, we aim to use human motion patterns for detecting pedestrians in a scene. Even though individuals have particular gaits, thier motion in general has inherent similarity and is invariant to clothing and appearance. We use dense optical flow map as a measure of motion in the scene and learn a discriminative classifier using "selected" eigen vectors as features. In fact, a cascade of such classifiers is learnt which not only provides speed-up but also improves discriminative power of the classifier. For more details, please see the Pedestrian Detection Project Page . |
 |
CMUcam3: Low cost embedded vision system
CMUcam3 is a cheap, low powered embedded system providing vision capabilities in an easy to use environment. With the help of open source libraries and example codes, it can be programmed in C for simple applications like color tracking or more complex ones like face detection. For more details on the project, please see the CMUCam3 Project Page . |
|
This
website is maintained by Devi Parikh
|