Student Research Projects


Supervisor
Li Cheng, Ajunct Research Fellow (ANU) & Researcher, Statistical Machine Learning Program (NICTA)

These projects are suitable as:
summer research projects
honours thesis projects
 

Motion detection


Project Code: SML_21

How to detect moving objects automatically and in real time from, for example, a hand-held camera or a robot? You will have opportunity to explore exciting computer vision (for example, SIFT detector from UBC) and machine learning techniques for a sensible solution. In this project, you will also enjoy playing with these robots (www.webcam32.com/SRV_info.html).

It will involve coding in C/C++, and require good knowledge of linear algebra, and some working knowledge of image/video processing.

Supervision will be provided on a day-to-day basis.


Action recognition


Project Code: SML_22

Given a video sequence of human activities, it is easy for us to tell whether a person is walking, running, or dancing. The aim of this project is to teach a computer to recognize human actions, that is, we predefine a set of actions that reasonably cover over the possible human activities in the obtained and incoming video sequences. Then the computer will learn to recognize these actions from a new video sequence.

This turns out to be rather difficult for computer vision, mainly due to different viewing angles, various poses within and across the category of actions. Nevertheless it attracts a lot of attentions in recent years, see www.cs.berkeley.edu/projects/vision/action for a collection of related research papers.

In this project, you are expect to explore many interesting computer vision (eg. shape analysis) and machine learning (eg. probabilistic models) techniques.

The summer scholar is expected to be confident in C/C++ and python programming, as well as have appropriate mathematics training (linear algebra, probability and statistics). Some working knowledge of image/video processing will be beneficial.

The summer scholar will work closely with the supervisor on a day-to-day basis.

Novelty detection from video data


Project Code: SML_23


Modern technologies enable us to easily obtain tons of video sequences everyday. Usually the recorded scenarios are similar things that repeat over the time. For example, from a live traffic control camera, we might observe vehicles running along the lanes most of the time. However, there are occasions when certain vehicle break the traffic law, and these are the situations that drivers and traffic officers have to pay attention to. Human is capable of monitoring and figuring out these novel situations, however, no one can do this consistently 24 hours a day, 365 days a year. This project aims at enabling computer to automatically detect these situations.

This problem is very important to, eg., surveillance industry and have potentials for various applications. Many efforts have been taken over the years, the solutions are still far from being satisfactory. To have a flavor, here is one related paper:
---- O. Boiman, M. Irani,  Detecting Irregularities in Images and in Video.   International Conference on Computer Vision (ICCV), Beijing, 2005.

During this project, the summer student will have expose to various statistical machine learning methods, such as Bayesian method and kernel method, as well as recent image/video processing techniques.

The summer scholar is expected to be confident in C/C++ and python programming, and have appropriate mathematics training (linear algebra, probability and statistics). Some prior knowledge of image/video processing will be beneficial.

The summer scholar will work closely with the supervisor on a day-to-day basis.