Institute for Visual Computing Founded

January 2009

Marc Pollefeys has been Full Professor at the Department of Computer Science since 2007 where he leads the Computer Vision and Geometry Lab. In December 2008 be became head of the newly founded Institute for Visual Computing. An interview with Katja Abrahams.

You are the head of the newly founded Institute of Visual Computing. What is the reason for the foundation of this institute and what groups does it consist of?

The new institute combines computer vision and computer graphics, two research areas which have been co-existing at the Department of Computer Science and become more and more closely related over the last couple of years. The time was now ripe to found this institute since the Institute of Computational Science (ICOS) whose research activities had been including these activities had to be restructured for organizational matters and ceased to exist end of 2008.

The Computer Graphics Laboratory (CGL), headed by Prof. Markus Gross and the Applied Geometry Group (AGG), headed by Prof. Mark Pauly, both Computer Graphics groups and former ICOS members, now form the Institute of Visual Computing together with my Computer Vision and Geometry Laboratory (CVG). Apart from that, there will be collaborations within the department, e.g. with Prof. Joachim Buhmann’s Machine Learning Laboratory.

What is the difference between computer vision and computer graphics?

Both research areas developed independently from each other, however they complement each other perfectly. In computer vision we take images of objects, scenes and events, analyze them with the computer and generate models which describe what can be seen in the images. These models can be geometric 3D models, but can also include motion, appearance and semantic information. The biggest challenge in computer vision is the complexity of the real world. It is very hard for the computer to recognize and reconstruct certain objects.

Computer graphics, on the other hand, often starts from such models and generates digital images from them. In some sense both are each others inverse. To give an example, Prof. Mark Pauly and his group are interested in the dynamic 3D models we capture in our group to analyze how shapes deform and how they can be rendered. Generally speaking, a lot of applications, especially in medicine and robotics, require both techniques.

How will the curriculum be affected by the foundation of the new institute?

Regarding education we are already ahead of the foundation of the new institute. A Master specialization track in Visual Computing, combing computer vision and computer graphics, as well as machine learning, has already been existing for several years and we will strengthen and extend this program. Moreover, we will have very interesting student projects and theses for D-INFK students, as well as for students at D-MAVT and D-ITET. For example, one of my master students who received a scholarship within ETH's Excellence Scholarship & Opportunity Program works on a project to build a micro helicopter called PIXHAWK. Currently we are setting up an interdisciplinary student team for this project. A prototype and platform will be developed to test computer vision approaches on micro aerial vehicles in order to navigate and do all image processing onboard, leading to full autonomy. Our helicopter will also participate in international micro aerial vehicles competitions.

Will there be collaborations with other departments at ETH and the industry in terms of research?

We have a common European project with Roland Siegwart from D-MAVT on autonomous flying micro-aerial vehicles and I also expect there will be collaboration with the Computer Vision Laboratory at D-ITET. The Computer Graphics Laboratory (CGL) of Prof. Markus Gross also has a tradition of collaboration with other departments. Furthermore, we have a project together with the University of North Carolina at Chapel Hill and Honda R&D, enabling Honda’s humanoid robot ASIMO to explore environments on its own by using computer vision technology. There are other ongoing collaborations with companies such as Microsoft, Nokia and Nvidia. Both CVG and CGL also collaborate with the ETH spin-off LiberoVision. And there will of course be a close collaboration with the Disney Research Lab here at ETH which is currently recruiting its members.

Is there anything you would like to be improved to create perfect working conditions for the Institute of Visional Computing?

Unfortunately, at the moment the different groups are located in two buildings, the computer graphics groups are located in the IFW building while the computer vision group is at five minutes walk in the CAB building. We would like to have a common lab space for experiments and to share equipment, such as computers cameras and projectors. This would also make the exchange between the researchers of the different groups much easier. In fact, in CAB we are currently lacking adequate lab space and we hope that something can be done to solve this problem.

Groups within the Institute of Visual Computing

Computer Vision and Geometry Laboratory: http://www.cvg.ethz.ch

Computer Graphics Laboratory: http://graphics.ethz.ch

Applied Geometry Group: http://www.agg.ethz.ch

PIXHAWK Micro Air Vehicle Computer Vision Platform: http://www.pixhawk.ethz.ch