Graphics, Capture and Massively Parallel Computing
The main focus of our research interest is the development of digitization methods for all aspects of the real world. This includes capturing geometry, surface color and several other attributes, e.g. reflectance properties of objects. Most of our methods perform the reconstruction solely on photographs that have been taken under controlled conditions (such as constant lighting and known camera parameters). Our latest developments, however, are capable of dealing with photos taken under uncontrolled conditions, for example images from community photo collections such as flickr.
Beyond that, we are interested in massively parallel computing to exploit the full potential of nowadays graphics hardware to solve computationally expensive problems in various areas, for example visual computing and bio informatics.
mapMAP source code released
We just released the (CPU) source code of our high-performance MAP solver under a liberal open source license. To access code and the corresponding paper, head over to the mapMAP project page.
PFASUM: a substitution matrix from Pfam structural alignments
In: BMC Bioinformatics, 18(1), p.293, 2017
Decoupled Space and Time Sampling of Motion and Defocus Blur for Unified Rendering of Transparent and Opaque Objects
In: Proceedings of Pacific Graphics 2016, Okinawa, Japan, 2016
Rapid, Detail-Preserving Image Downscaling
In: SIGGRAPH Asia 2016, Macao, PR China, 2016
Fabian Langguth, Kalyan Sunkavalli, Sunil Hadap, Michael Goesele
In: Proceedings of the European Conference on Computer Vision (ECCV), 2016 go