Roman Shapovalov
CV | Blog | Facebook

Bayesian Methods Group and
Graphics & Media Lab,
Faculty of Computational Mathematics and Cybeernetics,
Moscow State University

Bld. 52, 1 Leninskie Gory,
Moscow, Russia
Office: 703

E-mail:

Roman V. Shapovalov

I am currently a PhD student in the Mathematical Prediction Department at Moscow State University. My research interests include (but are not limited to) machine learning, probabilistic graphical models, and their applications to computer vision. My advisor is Dmitry Vetrov. We collaborate with Pushmeet Kohli as a part of Microsoft Research programs in Russia.

My PhD thesis is about structural learning for semantic segmentation of images and point clouds. Specifically, I now study max-margin learning of CRFs with various loss functions, which is useful, inter alia, for learning from weakly supervised data. I also have some project ideas I would be happy to implement if only I had enough time, such as a simple home video surveillance system (which recognizes faces of friends coming by and greets them), or a content-based music recommendation system (although recent Google's interest to industrial-quality music analysis ruins my motivation). However, if you are ready for collaboration, feel free to contact me.

Selected publications

See the comprehensive list of my publications on Google Scholar Citations.

R. Shapovalov. “Structural learning methods for collective labelling problems ”. [In Russian] PhD Thesis, Lomonosov Moscow State University, 2014. [text coming soon]

R. Shapovalov, D. Vetrov, A. Osokin, P. Kohli. “Multi-utility Learning: Structured-output Learning with Multiple Annotation-specific Loss Functions”. ArXiv:1406.5910, 2014. [pdf]

R. Shapovalov, D. Vetrov, P. Kohli. “Spatial Inference Machines”. IEEE CVPR. Portland, 2013. [project, pdf, video]

A. Velizhev, R. Shapovalov, K. Schindler. “Implicit shape models for object detection in 3D point clouds”. ISPRS Congress. Melbourne, 2012. [pdf]

R. Shapovalov, A. Velizhev. “Cutting-Plane Training of Non-associative Markov Network for 3D Point Cloud Segmentation”. IEEE 3DIMPVT. Hangzhou, 2011. [project, pdf, slides-pdf, slides-pptx, poster-pdf, data]

R. Shapovalov, A. Velizhev, O. Barinova. “Non-associative Markov networks for 3D point cloud classification”. PCV. Paris, 2010. [project, pdf, slides-pptx by A.Velizhev]

R. Shapovalov. “Automated object detection in laserscanning data”. [In Russian] Masters Thesis, Lomonosov Moscow State University, 2010. [pdf, slides-pdf, slides-pptx]

Code

I have some code on github.

GML LidarK Library — the library for LIDAR data processing. Currently, only the indexing data structure is implemented. It allows performing spatial queries in 3D space. The code is C++, MATLAB wrapper is also available.

GML BOLT — the toolkit for on-line learning from imbalanced streams. It contains an on-line implementation of the Random Forest algorithm.

Non-research interests

I enjoy snowboarding, cycling, urban orienteering and editing Wikipedia. I also learn to play electric guitar. And those are only virtuous ones! ;)

© Roman Shapovalov, 25-Jun-2014. Powered by X10 Hosting