Learning for Visual Data Synthesis and Analysis Event as iCalendar

(Science Event Tags, Computer Science, Seminars)

21 March 2019

12 - 1pm

Venue: 303S-561

Location: City

Host: School of Computer Science

Cost: Free

timo_ropinski

Speaker: Timo Ropinski, head of the Visual Computing Group at Ulm University

Abstract 

In this talk, Timo Ropinski will present his recent advances in deep learning to synthesise and analyse visual data, such as images, volumes, and point clouds.

The presented approaches are loosely aligned along the classical computer graphics rendering pipeline, whereby both structured and unstructured data are handled.

He will first present concepts for learning in object space, i.e., directly on the data to be rendered. To realise different visual tasks, such as normal estimation and segmentation, he will discuss how Monte Carlo integration can be used to realise convolutions on point cloud data, which represents the meshes to be rendered.

Based on this unstructured learning approach, he will further show, how this technology can be modified to replace the conventional rendering process to generate shaded images.

Finally, he will discuss a structured learning approach, which enables him to revert the image synthesis process, i.e., generating a volumetric data set out of a synthesised image.

For all three approaches, he will discuss training data generation, network architectures, and the obtained testing results.

About our speaker

Timo Ropinski heads the Visual Computing Group at Ulm University. Before moving to Ulm, he was Professor in Interactive Visualization at Linkoping University in Sweden where he was heading the Scientific Visualization Group.

Timo Ropinski has received his Ph.D. in computer science in 2004 from the University of Munster, where he has also completed his Habilitation in 2009.

His research foci lie in the areas of computer graphics and visualisation. He is especially interested in interactive visual problem solving, multimodal and multivariate visualisation, volume rendering as well as volumetric illumination models, all with a special consideration of large data sets.