Martin R. Oswald received his “Vordiplom” degree (2003) and a “Diplom” degree (2007) in Computer Science from Dresden University of Technology. In 2005 he studied Machine Learning and Computer Vision at the University of Technology, Sydney. From 2005 to 2007 he worked at the Institute for Civil Engineering at Dresden University of Technology studying neural networks and their application to engineering problems. Subsequently, he studied reliability estimation methods in civil engineering and received a Master's degree in civil engineering from the University of Technology Federico Santa Maria, Valparaiso - Chile (2008). From January 2009 to August 2014 he was a Ph.D. student in the Computer Vision Group at Technische Universität München, Germany, headed by Professor Daniel Cremers.
Spatiotemporal Multi-view 3D Reconstruction, Video Processing, Single View Reconstruction, Flow Estimation, Image Segmentation, Variational Methods, Optimization Methods, Probabilistic Methods, Machine Learning, Artificial Intelligence.
I have been working on the following research projects:
Spatio-Temporal Multi-View 3D Reconstruction
Problem: Recover the 3D geometry of a scene from multiple synchronously captured videos.
The video below shows reconstruction results of our method with our proposed generalized connectivity constraints (published at ECCV 2014).
The following video shows results of our method which additionally estimates and integrates normal information (published at BMVC 2014)
The following video shows early results of our method (published at ICCV-4DMOD 2013).
For more information, see the related publication.
|Conference and Workshop Papers|
|Generalized Connectivity Constraints for Spatio-temporal 3D Reconstruction , In European Conference on Computer Vision (ECCV), 2014. [bib] [pdf] [video]|
|Surface Normal Integration for Convex Space-time Multi-view Reconstruction , In British Machine Vision Conference (BMVC), 2014. [bib] [pdf] [video]|
|Spatial and Temporal Interpolation of Multi-View Image Sequences , In German Conference on Pattern Recognition (GCPR), volume 36, 2014. [bib] [pdf] [video]|
|A Convex Relaxation Approach to Space Time Multi-view 3D Reconstruction , In ICCV Workshop on Dynamic Shape Capture and Analysis (4DMOD), 2013. [bib] [pdf]|
Single-View 3D Reconstruction
Problem: Recover the 3D geometry of an object from a single input image. For example:
The following video shows several results generated by our single-view reconstruction tool.
For more details have a look at the project page.