archive-at.com » AT » T » TUGRAZ.AT

Total: 196

Choose link from "Titles, links and description words view":

Or switch to "Titles and links view".
  • Learning, Recognition, and Surveillance @ ICG
    Navigation Home Members Research Publications Projects Downloads Partners Password protected resource Please enter credentials Login Password Copyright 2010 ICG

    Original URL path: http://lrs.icg.tugraz.at/research/aflw/downloads.php (2016-02-14)
    Open archived version from archive


  • demo_lfw_sift
    from automatically labeled bags of faces In Proc European Conf on Computer Vision 2010 clc clear all close all DATA OUT DIR fullfile dataOut cvpr lfw sift run toolbox init load fullfile DATA OUT DIR lfw sifts mat PARAMS params numCoeffs 100 params saveDir fullfile DATA OUT DIR all params title LFW SIFT LEARNING ALGORITHMS USED FOR CROSS VALIDATION algorithms for pairwise training pair metric learn algs LearnAlgoKISSME LearnAlgoMahal LearnAlgoMLEuclidean

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/demo_lfw_sift.html (2016-02-14)
    Open archived version from archive

  • demo_toycars
    inrialpes fr people nowak similarity index html Features HSV Lab histograms and LBPs 16 to describe color and texture of the non overlapping blocks size 30x30 The global image descriptor is a concatenation of the local ones Using PCA the descriptor is projected onto a 50 dimensional subspace INIT LOAD clc clear all close all run toolbox init DATA OUT DIR fullfile dataOut cvpr toy car lear load fullfile DATA OUT DIR toycars features mat PARAMS params pca numDims 50 we project onto the first 50 PCA dim params svm liblinear options B 10 s 2 c 1 params svm smoothing 0 params saveDir fullfile DATA OUT DIR out VALIDATE AND PLOT ROC CURVES algorithms for pairwise training pair metric learn algs LearnAlgoKISSME LearnAlgoMahal LearnAlgoMLEuclidean LearnAlgoSVM params svm LearnAlgoITML LearnAlgoLDML algorithms to train with class labels metric learn algs LearnAlgoLMNN DO VALIDATION ACCORDING TO TOYCARS PROTOCOL ds CrossValidatePairs struct pair metric learn algs pairs ux 1 params pca numDims idxa idxb ds CrossValidatePairs ds metric learn algs pairs ux 1 params pca numDims idxa idxb ToyCarPairsToLabels EVALUATION we evaluate only on the test set fold 2 train set fold 1 ignore rocPlot evalData pairs logical pairs training ds 2 params

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/demo_toycars.html (2016-02-14)
    Open archived version from archive

  • demo_viper
    HSV Lab histograms and LBPs 16 to describe color and texture of the overlapping blocks size 8x16 and stride of 8x8 pixels The image descriptor is a concatenation of the local ones Using PCA the descriptor is projected onto a 34 dimensional subspace clc clear all close all DATA OUT DIR fullfile dataOut cvpr viper run toolbox init m Set up parameters params numCoeffs 34 dimensionality reduction by PCA to 34 dimension params N 632 number of image pairs 316 to train 316 to test params numFolds 10 number of random train test splits params saveDir fullfile DATA OUT DIR all params pmetric 0 Load Features load fullfile DATA OUT DIR viper features mat Cross validate over a number of runs pair metric learn algs LearnAlgoKISSME params LearnAlgoMahal LearnAlgoMLEuclidean LearnAlgoITML LearnAlgoLDML LearnAlgoLMNN ds CrossValidateViper struct pair metric learn algs ux 1 params numCoeffs idxa idxb params Plot Cumulative Matching Characteristic CMC Curves names fieldnames ds for nameCounter 1 length names s ds names nameCounter ms names nameCounter cmc cat 1 s cmc params N 2 ms names nameCounter roccolor s 1 roccolor end h figure names fieldnames ms for nameCounter 1 length names hold on plot median ms names nameCounter

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/demo_viper.html (2016-02-14)
    Open archived version from archive

  • demo_pubfig
    CAVE databases pubfig clc clear all close all run toolbox init DATA OUT DIR fullfile dataOut cvpr pubfig load fullfile DATA OUT DIR pubfig attributes mat PREPROCESSING params pca numCoeffs 65 params title PUBFIG ATTRIBUTES params mu 0 smothing parameters see below params sigma 1 gaussian weighting see 13 for details X attributes g normpdf X 0 5 params mu params sigma X X g dimensionality reduction ux u m

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/demo_pubfig.html (2016-02-14)
    Open archived version from archive

  • LearnAlgoKISSME
    matrices Default 1 p lambda Influence of the dissimilar pairs Range 0 to 1 default 1 return obj object handle to instance of class LearnAlgoKISSME if nargin 1 p struct end if isfield p lambda p lambda 1 end if isfield p roccolor p roccolor g end if isfield p pmetric p pmetric 1 end obj p p check obj end function bool check obj bool check obj Checks if all dependencies are satisfied bool exist SOPD 0 if bool className class obj fprintf Sorry s not available n className end obj available bool end function s learnPairwise obj X idxa idxb matches s learnPairwise obj X idxa idxb matches Learn from pairwise equivalence labels parameters obj instance of class LearnAlgoKISSME X input matrix each column is an input vector DxN 2 N is the number of pairs D is the feature dimensionality idxa index of image A in X 1xN idxb index of image B in X 1xN matches matches defines if a pair is similar 1 or dissimilar 0 return s Result data struct s M Trained quadratic distance metric s t Training time in seconds s p Used parameters see LearnAlgoKISSME properties for details s s learnAlgo class handle to obj s roccolor line color for ROC curve default g if obj available s struct return end idxa double idxa idxb double idxb KISS Metric Learning CORE ALGORITHM tic Eqn 12 sum of outer products of pairwise differences similar pairs normalized by the number of similar pairs covMatches SOPD X idxa matches idxb matches sum matches Eqn 13 sum of outer products of pairwise differences dissimilar pairs normalized by the number of dissimilar pairs covNonMatches SOPD X idxa matches idxb matches sum matches t toc tic Eqn 15 16 s M inv covMatches obj p lambda inv

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/LearnAlgoKISSME.html (2016-02-14)
    Open archived version from archive

  • install3dpartylibs
    ismac run fullfile LMNN DIR install m else run fullfile LMNN DIR installWINDOWS m end clc clc fprintf Installing LMNN succeeded n pause 0 5 catch ME clc fprintf Warning Installing LMNN failed n pause 2 end Information Theoretic Metric Learning algorithm ITML License na Url http www cs utexas edu pjain itml try url http www cs utexas edu pjain itml download itml 1 2 tar gz untar url LIB DIR clc fprintf ITML installation succeeded n pause 0 5 catch ME clc fprintf Warning ITML installation failed n pause 2 end Logistic Discriminant based Metric Learning LDML License na Url http lear inrialpes fr people guillaumin code php mildml try url http lear inrialpes fr people guillaumin code MildML 0 1 tar gz untar url LIB DIR clc fprintf LDML installation succeeded n pause 0 5 catch ME clc fprintf Warning LDML installation failed n pause 2 end LIBLINEAR Linear SVM License New BSD License Url http www csie ntu edu tw cjlin liblinear try url http www csie ntu edu tw cjlin cgi bin liblinear cgi http www csie ntu edu tw cjlin liblinear zip unzip url LIB DIR if strcmp computer PCWIN64 LIB LIN DIR dir

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/install3dpartylibs.html (2016-02-14)
    Open archived version from archive

  • demo_lfw_attributes
    face recognition in unconstrained environments Technical Report 07 49 University of Massachusetts Amherst 2007 See also http vis www cs umass edu lfw Features 13 N Kumar A C Berg P N Belhumeur and S K Nayar Attribute and Simile Classifiers for Face Verification In Proc IEEE Intern Conf on Computer Vision 2009 See also http www cs columbia edu CAVE databases pubfig clc clear all close all DATA OUT DIR fullfile dataOut cvpr lfw attributes run toolbox init load fullfile DATA OUT DIR lfw attributes mat algorithms for pairwise training pair metric learn algs LearnAlgoKISSME LearnAlgoMahal LearnAlgoMLEuclidean LearnAlgoITML uncomment here to enable ITML LearnAlgoLDML uncomment here to enable LDML LearnAlgoSVM uncomment here to enable SVM algorithms to train with class labels metric learn algs LearnAlgoLMNN uncomment here to enable LMNN PREPROCESSING params pca numCoeffs 65 we project onto the first 65 PCA dim params saveDir fullfile DATA OUT DIR out params title LFW ATTRIBUTES params mu 0 smothing parameters see below params sigma 1 gaussian smoothing of the attribute features see 13 for details X attributes g normpdf X 0 5 params mu params sigma X X g ux u m applypca X DO CROSS VALIDATION ACCORDING TO LFW

    Original URL path: http://lrs.icg.tugraz.at/research/kissme/html/demo_lfw_attributes.html (2016-02-14)
    Open archived version from archive