Research

Reseach interest: Machine Learning, Computer Vision, Statistical Learning/Analysis

Current research: Object recognition, region segmentation, and image parsing for aerial/satellite images

Academic advisor: Song-chun Zhu

Previous Research and Course Projects:

1.    Master Thesis Research Project (Advisor: Prof. James J. Little and Prof. Nando de Freitas): Robust Visual Tracking for Multiple Targets

This work is published in ECCV 2006

We address the problem of robust multi-target tracking within the application of hockey player tracking. The particle filter technique is adopted and modified to fit into the multi-target tracking framework. A rectification technique is employed to find the correspondence between the video frame coordinates and the standard hockey rink coordinates so that the system can compensate for camera motion and improve the dynamics of the players. A global nearest neighbor data association algorithm is introduced to assign boosting detections to the existing tracks for the proposal distribution in particle filters. The mean-shift algorithm is embedded into the particle filter framework to stabilize the trajectories of the targets for robust tracking during mutual occlusion. Experimental results show that our system is able to automatically and robustly track a variable number of targets and correctly maintain their identities regardless of background clutter, camera motion and frequent mutual occlusion between targets.

eccv1.JPG            Comparison between the old system and the new system

eccv3.JPG        More results

1.    Final project of CPSC 525 Image Understanding II (Instructor: Prof. David Lowe, UBC): Invariant Local Features for Face Detection

In the domain of object recognition, the SIFT feature is known to be a very successful local invariant feature. The performance of the recognition task using SIFT features is very robust and also can be done in real-time. This project present an approach that adopt the SIFT feature for the task of face detection. A feature database is created for the detection of generic face features and a model fitting algorithm is used to resolve the scale, orientation and position of the faces in images. Because of the advantages brought by the SIFT features, the system can easily detect faces of any scale and rotation. A test is done using 16 images with 200 faces in total. The result shows a detection rate of 72.5% with 46 false detections.

image001.jpg

2.    Final project of CPSC 522 Artificial Intelligence II (Instructor: Prof. David Poole, UBC): Reinforcement Learning for Mobile Robots with Continuous Space

It is very tedious to program a mobile robot in the real world. Also, it is difficult for the programmed mobile robot to get adapted to a new environment. Reinforcement learning provides a proper mechanism for the mobile robot to learn how to accomplish a task in any given environment with very little work for programming. However, traditional reinforcement learning methods only assume discrete state and action space, which is not applicable for mobile robots in the real world. This project simulated a mobile robot with continuous states and discrete actions using a safe approximation of the value function to learn the optimal policy. Experiment result shows that learning can be very successful and efficient with bootstrapped information provided by human controls.

image004.jpg

3.    Final project of CPSC 540 Machine Learning (Instructor: Prof. Nando de Freitas, UBC): Image Retrieval Using Boosting Algorithm

This project uses an offline learning algorithm to get a highly efficient classifier for online image retrieval. The boosting algorithm is adopted for the learning process. It chooses a small number (10 in this project) of highly selective features from a very large feature set (there are totally over 45,000 features in the set in this project) and combine them together to get a strong classifier for retrieval. The computation of such strong classifier is very fast so that it can be applied to image retrieval on a very large dataset.

image015.gif

4.    Final project of CPSC 508 Distributed System (Instructor: Prof. Norman C. Hutchinson, UBC): Distributed File Sharing based on Peer-to-Pear Architecture

The classical Server-Client architecture has dominated the world of computer networks for decades. However, with the booming of the information and multimedia technology, the demand for large file sharing has increased drastically. The classical Server-Client model suffers from the overwhelmingly large load of download and upload of files. The Peer-to-Peer network architecture has solved this problem by distributing the files onto the computers on the network. This project implemented a file sharing system that supports P2P file sharing among multiple computers based on the architecture used in the Bit-Torrent.

5.    Undergraduate Thesis: Face Detection System Design and Implementation based on Ada-Boost

The purpose of this thesis is to use the latest real-time feature detection algorithms, developed by Paul Viola, to implement a fast human face detection system. Through the experiment, we try to get the optimal parameters to construct a human face detector with high detection speed and detection rate. Other than the implementation of the algorithms developed by Paul, we also developed a method for image data preprocess of the training system, a quick image scanning algorithm, an algorithm for simple classifier initialization and its optimization. Finally, we will have an analysis on the issue of feature selection with different system parameters and the detection result of the whole system.

Course work

Spring, 2006:
STATS 200C Large Sample Theory 
STATS 201C Advance Modeling & Data Mining
STATS 202C Markov Chain Monte Carlo & Optimization Algorithms
STATS 232B Computing & Inference in Computer Vision

Winter, 2006:
STATS 200B Applied Probability 
STATS 201B Regression Analysis
STATS 232A Modeling & Learning in Computer Vision

Fall, 2005:
STATS 200A Applied Probability 
STATS 201A Research Design & Data Management
STATS M231 Pattern Recognition

Winter Session Term 1, 2005:
CPSC 532c Probabilistic graphical models

Winter Session Term 2, 2004:
CPSC 522 Artificial Intellegence II
CPSC 525 Image Understanding II
Machine Learning Reading Group
Reinforcement Learning Reading Group

Winter Session Term 1, 2003:
CPSC 508 Distributed Systems
CPSC 540 Machine Learning
CPSC 590 Research Methods in Computer Science

TA work

 

Winter, 2006:
TA of Biological Statistics
Course instructor: Mark Hansen

My obligations: Marking

 

Winter Session Term 2, 2004:
TA of CPSC 417, Computer Communication
Course instructor: Alan Wagner
My obligations:
Giving tutorial on Socket Programming in C and low level bit operation in C
Answer questions in the course newsgroup
Marking

 

Winter Session Term 1, 2003:
TA of CPSC 315, Operating Systems
Course instructor: Ed Knorr
My obligations:
Hold office hours
Answer questions on WebCT
Marking