001/**
002 * Copyright (c) 2011, The University of Southampton and the individual contributors.
003 * All rights reserved.
004 *
005 * Redistribution and use in source and binary forms, with or without modification,
006 * are permitted provided that the following conditions are met:
007 *
008 *   *  Redistributions of source code must retain the above copyright notice,
009 *      this list of conditions and the following disclaimer.
010 *
011 *   *  Redistributions in binary form must reproduce the above copyright notice,
012 *      this list of conditions and the following disclaimer in the documentation
013 *      and/or other materials provided with the distribution.
014 *
015 *   *  Neither the name of the University of Southampton nor the names of its
016 *      contributors may be used to endorse or promote products derived from this
017 *      software without specific prior written permission.
018 *
019 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
020 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
021 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
022 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
023 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
024 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
025 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
026 * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
027 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
028 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
029 */
030package org.openimaj.demos;
031
032import java.io.File;
033import java.io.IOException;
034import java.util.ArrayList;
035import java.util.Collections;
036import java.util.List;
037
038import org.openimaj.feature.local.list.MemoryLocalFeatureList;
039import org.openimaj.image.feature.local.aggregate.FisherVector;
040import org.openimaj.image.feature.local.keypoints.FloatKeypoint;
041import org.openimaj.image.feature.local.keypoints.Keypoint;
042import org.openimaj.math.matrix.algorithm.pca.ThinSvdPrincipalComponentAnalysis;
043import org.openimaj.math.statistics.distribution.MixtureOfGaussians;
044import org.openimaj.ml.gmm.GaussianMixtureModelEM;
045import org.openimaj.ml.gmm.GaussianMixtureModelEM.CovarianceType;
046import org.openimaj.util.array.ArrayUtils;
047
048import Jama.Matrix;
049
050public class FisherTesting {
051        public static void main(String[] args) throws IOException {
052                final List<MemoryLocalFeatureList<FloatKeypoint>> data = new ArrayList<MemoryLocalFeatureList<FloatKeypoint>>();
053                final List<FloatKeypoint> allKeys = new ArrayList<FloatKeypoint>();
054
055                for (int i = 0; i < 100; i++) {
056                        final MemoryLocalFeatureList<FloatKeypoint> tmp = FloatKeypoint.convert(MemoryLocalFeatureList.read(
057                                        new File(String.format("/Users/jsh2/Data/ukbench/sift/ukbench%05d.jpg", i)), Keypoint.class));
058                        data.add(tmp);
059                        allKeys.addAll(tmp);
060                }
061
062                Collections.shuffle(allKeys);
063
064                final double[][] sample128 = new double[1000][];
065                for (int i = 0; i < sample128.length; i++) {
066                        sample128[i] = ArrayUtils.convertToDouble(allKeys.get(i).vector);
067                }
068
069                System.out.println("Performing PCA " + sample128.length);
070                final ThinSvdPrincipalComponentAnalysis pca = new ThinSvdPrincipalComponentAnalysis(64);
071                pca.learnBasis(sample128);
072                final double[][] sample64 = pca.project(new Matrix(sample128)).getArray();
073
074                System.out.println("Projecting features");
075                for (final MemoryLocalFeatureList<FloatKeypoint> kpl : data) {
076                        for (final FloatKeypoint kp : kpl) {
077                                kp.vector = ArrayUtils.convertToFloat(pca.project(ArrayUtils.convertToDouble(kp.vector)));
078                        }
079                }
080
081                System.out.println("Learning GMM " + sample64.length);
082                final GaussianMixtureModelEM gmmem = new GaussianMixtureModelEM(512, CovarianceType.Diagonal);
083                final MixtureOfGaussians gmm = gmmem.estimate(sample64);
084
085                final double[][] v1 = gmm.logProbability(new double[][] { sample64[0] });
086                final double[][] v2 = MixtureOfGaussians.logProbability(new double[][] { sample64[0] }, gmm.gaussians);
087
088                System.out.println("Done");
089
090                // for (int i = 0; i < 512; i++) {
091                // System.out.println(gmm.gaussians[i].getMean().get(0, 0) + "," +
092                // gmm.gaussians[i].getMean().get(0, 1));
093                // }
094
095                final FisherVector<float[]> fisher = new FisherVector<float[]>(gmm, true, true);
096
097                final List<FloatKeypoint> kpl = allKeys.subList(0, 26000);
098                final long t1 = System.currentTimeMillis();
099                fisher.aggregate(kpl).asDoubleVector();
100                final long t2 = System.currentTimeMillis();
101                System.out.println(t2 - t1);
102
103                // int i = 0;
104                // final double[][] fvs = new double[5][];
105                // for (final MemoryLocalFeatureList<FloatKeypoint> kpl : data) {
106                // final long t1 = System.currentTimeMillis();
107                // fvs[i++] = fisher.aggregate(kpl).asDoubleVector();
108                // final long t2 = System.currentTimeMillis();
109                // System.out.println(t2 - t1);
110                //
111                // if (i == 5)
112                // break;
113                // }
114                //
115                // final ThinSvdPrincipalComponentAnalysis pca2 = new
116                // ThinSvdPrincipalComponentAnalysis(128);
117                // pca2.learnBasis(fvs);
118        }
119}