computer science

Natural User Interfaces (Submitted)

Currently working on a natural user interface.. The aim is to develop a toy world that can be manipulated in a natural manner. This project is under the supervision of Geoff Wyvill and Alistair Knott.

Towards 3D 'Photocopying' (Submitted)

I have been interested for a number of years in the Reprap project , a project to design and build a hobbyist 3d printer that can print its own parts. A 3d printer is a great tool if you have the expertise to make good use of 3d modelling tools but for my PhD I wanted to see if I could give the average person a simple way to duplicate an object by deducing a 3d model from a set of digital photographs taken with a normal handheld digital camera. The software that has been created as part of my PhD is open-source, available on SourceForge, and was designed to be part of the Reprap software tool-chain. My supervisors are Geoff Wyvill in Computer Science and Holger Regenbrecht in Infomation Science, both at Otago University.

Indoor Navigation for the Blind

I am a Ph.D. student at Otago University, the Computer Science Department. I am supervised by Associate Professor Brendan McCane. My co-supervisor is Geoff Wyvill. I am currently in Vision/ Graphics research lab. I am on a PhD scholarship and started my PhD in 2010. My research interests include computer vision, distributed systems and databases. Currently I am working on 'Blind People Navigation in Indoor' in my PhD. I am looking to make an application for blind which can help them to navigate inside the buildings. I got my Master degree in Computer Software Engineering from Arid and National University of Sciences and Technology, Pakistan in 2007. Before coming to Otago, I was working as a Lecturer in my home university since 2008. It has been a wonderful experience in New Zealand. For more information please visit my Personal Homepage.

Bridging the Semantic Gap

At present there is no effective system that can identify objects in an image. This problem is known as semantic gap problem. I will try to solve this problem by scaling up the size of the images collection and then finding semantic information presented in them. As images containing similar objects produce similar features which can be used to retrieve image from a collection and this is how the Bag-of-Words method works. Just as a child learns something through repeated exposures of an object so as a computer can. So for this technique to be effective each object should have many exposures in the collection.

Surface Matching Applied to Medical Imaging

Hamza Bennani from Morocco has joined in June 2011 the graphics lab at University of Otago, Dunedin, as a PhD Student on surface matching applied to medical imaging. He is a Computer Science and Applied Mathematics Engineer graduating from applied mathematics and computer science department at ENSEEIHT Toulouse, France. In 2009/2010, Hamza was an ERASMUS student in TU-Darmstadt, Germany. During the same period he worked at the Innovation and Integration Systems department at T-Systems, Darmstadt where he worked on machine learning and computer vision techniques. This work was the support of his masters thesis, involving the conception of an application for mobile phones able to recognize 2D Barcode on real time using video streaming. His current work is really exciting, at the interface between human health and computer science. He enjoys creating transdisciplinary links and creating tools useful for everyone.

Real Time Robust Eye Tracking in the Watching Window

Maria is working on eye tracking for the 'Watching Window' project, which presents a customised 3D view of an object to a user. This view changes as the user moves their head, giving an illusion of depth via motion parallax.

Manifold and Non-Mainifold Learning for Computer Vision

Xiping is investigating dimension estimation and manifold learning techniques to represent low-dimensional sub-spaces in high-dimensional data.

Co-operative Tracking in Distributed Camera Networks

Jordan is interested in tracking across multiple views, when the views don't overlap. This involves learning how activity in one view relates across time to another view.

Texture Generation

Thomas is investigating various mathematical models of noise with applications to texture generation.

Segmentation of Neuronal MRI Images

Ben is working on the problem of segmenting Neuronal MRI images for medical analysis.