Victor Lempitsky

Виктор Лемпицкий

 		email
+44 1223 479 925
I am a postdoc researcher in Computer Vision group at Microsoft Research Cambridge working on computer vision and pattern recognition.
I did my undergraduate studies and obtained my PHD degree at the Department of Mathematics and Mechanics at Moscow State University.
I was also a visiting PHD researcher at Vision Research Group at the University of Western Ontario.
 		  photo 

 

Selected Publications 

 
  • Fusion Moves for Markov Random Field Optimization. PDF
    V. Lempitsky, C. Rother, S. Roth, and A. Blake
    Technical Report MSR-TR-2009-60
    to appear in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI).
  • Image Segmentation with A Bounding Box Prior.
    V. Lempitsky, P. Kohli, and C. Rother
    IEEE International Conference on Computer Vision (ICCV)    , Kyoto, 2009 (to appear, oral presentation).
  • Class-Specific Hough Forests for Object Detection. PDF
    J. Gall and V. Lempitsky
    IEEE Computer Vision and Pattern Recognition (CVPR)    , Miami, 2009 (oral presentation).
  • Random Forest Classification for Automatic Delineation of Myocardium in Real-time 3D Echocardiography. PDF
    V. Lempitsky, M. Verhoek, A. Noble, and A. Blake
     Functional Imaging and Modelling of the Heart (FIMH), Nice, 2009 (oral presentation, best paper award).
  • Image Segmentation by Branch-and-Mincut. PDF, PDF (of the extended tech. report), PPT, code below
         V. Lempitsky, A. Blake, and C. Rother
    European Conference on Computer Vision (ECCV)    , Marseille, 2008 (oral presentation).
  • FusionFlow: Discrete-Continuous Optimization for Optical Flow Estimation. PDF
    V. Lempitsky, S. Roth, and C. Rother
    IEEE Computer Vision and Pattern Recognition (CVPR)    , Anchorage, 2008.
  • LogCut - Efficient Graph Cut Optimization for Markov Random Fields. PDF
    V. Lempitsky, C. Rother, and A. Blake
    IEEE International Conference on Computer Vision (ICCV)    , Rio de Janeiro, 2007.
  • Global Optimization for Shape Fitting. PDF, code below
    V. Lempitsky and Y. Boykov
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Minneapolis,     2007.
  • Seamless Mosaicing of Image-Based Texture Maps. PDF
    V. Lempitsky and D. Ivanov
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Minneapolis, 2007    .
  • Optimizing binary MRFs via extended roof duality. PDF
    C. Rother, V. Kolmogorov, V. Lempitsky, and M. Szummer
    IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Minneapolis, 2007     (oral presentation).
  • Form Photohulls to Photoflux Optimization.  PDF (of the extended tech. report)
    Y. Boykov and V. Lempitsky
    British Machine Vision Conference (BMVC), Edinburgh, 2006.
  • Oriented visibility for Multiview Reconstruction. PDF
    V. Lempitsky, Y. Boykov, and D. Ivanov
    European Conference on Computer Vision (ECCV), Graz, 2006    .

Code
(available for research purposes only)

The C++ code for the touch-expand algorithm as described in the "Global Optimization for Shape Fitting" paper. It can be used for the shape-from-weakly oriented points tasks as well as for other minimum cut problems on 3D voxel grids, where the t-links are sparse or have large regions of the same sign. The code outputs the binary subdivision of the voxel volume into interior/exterior. A simple MATLAB script for the surface extraction is then provided (an advanced surface extraction algorithm is not included at the moment). 

The C++ code for the Branch-and-Mincut framework as described in the "Image Segmentation by Branch-and-Mincut" ECCV 2008 paper. The code contains the prototype class for the framework as described in section 3, as well as the sample implementation for the discrete Chan-Vese functional minimization as described in section 5.1.