Deep multi-scale video prediction beyond mean square error - Torch (lua) Code

  • Paper: Michael Mathieu, Camille Couprie, Yann LeCun "Deep multi-scale video prediction beyond mean square error", ICLR 2016
  • Description: Check out the videos produced using this code!
  • Keywords: Convolutional networks, predictive learning
  • Download: Torch code

    • Dual Constrained TV-based Regularization on graphs - Matlab Code

  • Paper: Camille Couprie, Leo Grady, Laurent Najman, Jean-Christophe Pesquet, Hugues Talbot "Dual Constrained TV-based Regularization on graphs", SIAM Journal on Imaging Science, 2013
  • Description: This software
  • Keywords: Proximal methods, inverse problems, image denoising, convex optimization, image restoration, mesh smoothing, discrete calculus.
  • Download: Matlab code
    •  

       

    		 
     

               (a) mesh denoising                                (b) non-local image denoising                  

    		 

    The Power Watershed Algorithm C/C++ Code

  • Paper: Camille Couprie, Leo Grady, Laurent Najman and Hugues Talbot, "Power Watersheds: A New Image Segmentation Framework Extending Graph Cuts, Random Walker and Optimal Spanning Forest", ICCV'09, 2009
  • Description: This software provides implementation of three algorithms for 2D and 3D image segmentation. The algorithms implemented are Prim, and Kruskal's algorithms for Maximum Spanning Forest computation, and the PowerWatershed algorithm (q=2) described in the above article.
  • Keywords: image segmentation, combinatorial optimization, mathematical morphology
  • Download: C/C++ Code available at https://sourceforge.net/projects/powerwatershed
    •  

               

    		 
     

               (a) Input seeds                        (b)Kruskal's MaxSF algorithm         (c) Prim's MaxSF algorithm          (d) Power Watershed with q=2         

    		 

    Causal graph-based video segmentation C/C++ Code

  • Paper: Camille Couprie, Clement Farabet, Yann Lecun "Causal graph-based video segmentation ", 2012
  • Description: This software takes a sequence of images extracted from a video and produces a temporally consistent segmentation with a random color assigned to each region. If semantic predictions are available, it exploits them to compute a temporally consistent semantic segmentation.
  • Keywords: semantic segmentation, scene parsing, graph matching, mathematical morphology
  • Download: C/C++ Code with examples of data (NYU-Scene dataset)
    •  

               

     

               

    		 
     

       Top row: Segmentation using Felzenszwalb and Huttenlocher approach. Bottom row: Our temporal segmentation result   

    		 


    Surface reconstruction C/C++ Code

  • Paper: Camille Couprie, Xavier Bresson, Laurent Najman, Hugues Talbot, and Leo Grady "Surface reconstruction using Power Watershed", ISMM'11
  • Description: This software provides implementation of the Power Watershed algorithm (q=2) for Surface reconstruction from noisy point clouds, described in the above article.
  • Keywords: semantic segmentation, scene parsing, graph matching, mathematical morphology
  • Download: Code
    • Design downloaded from free website templates.