View Javadoc

1   /**
2    * Copyright (c) 2011, The University of Southampton and the individual contributors.
3    * All rights reserved.
4    *
5    * Redistribution and use in source and binary forms, with or without modification,
6    * are permitted provided that the following conditions are met:
7    *
8    *   * 	Redistributions of source code must retain the above copyright notice,
9    * 	this list of conditions and the following disclaimer.
10   *
11   *   *	Redistributions in binary form must reproduce the above copyright notice,
12   * 	this list of conditions and the following disclaimer in the documentation
13   * 	and/or other materials provided with the distribution.
14   *
15   *   *	Neither the name of the University of Southampton nor the names of its
16   * 	contributors may be used to endorse or promote products derived from this
17   * 	software without specific prior written permission.
18   *
19   * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
20   * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
21   * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
22   * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
23   * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
24   * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25   * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
26   * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
27   * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
28   * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29   */
30  package org.openimaj.demos;
31  
32  import java.io.File;
33  import java.io.IOException;
34  import java.util.Arrays;
35  
36  import org.openimaj.feature.FloatFV;
37  import org.openimaj.feature.local.list.MemoryLocalFeatureList;
38  import org.openimaj.image.feature.dense.gradient.dsift.FloatDSIFTKeypoint;
39  import org.openimaj.image.feature.local.aggregate.FisherVector;
40  import org.openimaj.math.matrix.algorithm.pca.PrincipalComponentAnalysis;
41  import org.openimaj.math.matrix.algorithm.pca.ThinSvdPrincipalComponentAnalysis;
42  import org.openimaj.math.statistics.distribution.DiagonalMultivariateGaussian;
43  import org.openimaj.math.statistics.distribution.MixtureOfGaussians;
44  import org.openimaj.math.statistics.distribution.MultivariateGaussian;
45  import org.openimaj.util.array.ArrayUtils;
46  
47  import Jama.Matrix;
48  
49  import com.jmatio.io.MatFileReader;
50  import com.jmatio.io.MatFileWriter;
51  import com.jmatio.types.MLArray;
52  import com.jmatio.types.MLDouble;
53  import com.jmatio.types.MLSingle;
54  import com.jmatio.types.MLStructure;
55  
56  /**
57   * 
58   * @author Sina Samangooei (ss@ecs.soton.ac.uk)
59   */
60  public class FVFWCheckPCAGMM {
61  
62  	private static final String GMM_MATLAB_FILE = "/Users/ss/Experiments/FVFW/data/gmm_512.mat";
63  	private static final String PCA_MATLAB_FILE = "/Users/ss/Experiments/FVFW/data/PCA_64.mat";
64  	private static final String[] FACE_DSIFTS = new String[] {
65  			"/Users/ss/Experiments/FVFW/data/Aaron_Eckhart_0001-pdfsift.bin"
66  	};
67  
68  	public static void main(String[] args) throws IOException {
69  		final MixtureOfGaussians mog = loadMoG(new File(GMM_MATLAB_FILE));
70  		final PrincipalComponentAnalysis pca = loadPCA(new File(PCA_MATLAB_FILE));
71  		final FisherVector<float[]> fisher = new FisherVector<float[]>(mog, true, true);
72  		for (final String faceFile : FACE_DSIFTS) {
73  			final MemoryLocalFeatureList<FloatDSIFTKeypoint> loadDSIFT = MemoryLocalFeatureList.read(new File(faceFile),
74  					FloatDSIFTKeypoint.class);
75  			projectPCA(loadDSIFT, pca);
76  
77  			final FloatFV fvec = fisher.aggregate(loadDSIFT);
78  			System.out.println(String.format("%s: %s", faceFile, fvec));
79  			System.out.println("Writing...");
80  
81  			final File out = new File(faceFile + ".fisher.mat");
82  			final MLArray data = toMLArray(fvec);
83  			new MatFileWriter(out, Arrays.asList(data));
84  		}
85  	}
86  
87  	private static MLArray toMLArray(FloatFV fvec) {
88  		final MLDouble data = new MLDouble("fisherface", new int[] { fvec.values.length, 1 });
89  		for (int i = 0; i < fvec.values.length; i++) {
90  			data.set((double) fvec.values[i], i, 0);
91  		}
92  		return data;
93  	}
94  
95  	private static void projectPCA(
96  			MemoryLocalFeatureList<FloatDSIFTKeypoint> loadDSIFT,
97  			PrincipalComponentAnalysis pca)
98  	{
99  		for (final FloatDSIFTKeypoint kp : loadDSIFT) {
100 			kp.descriptor = ArrayUtils.convertToFloat(pca.project(ArrayUtils.convertToDouble(kp.descriptor)));
101 			final int nf = kp.descriptor.length;
102 			kp.descriptor = Arrays.copyOf(kp.descriptor, nf + 2);
103 			kp.descriptor[nf] = (kp.x / 125f) - 0.5f;
104 			kp.descriptor[nf + 1] = (kp.y / 160f) - 0.5f;
105 		}
106 		loadDSIFT.resetVecLength();
107 	}
108 
109 	static class LoadedPCA extends ThinSvdPrincipalComponentAnalysis {
110 
111 		public LoadedPCA(Matrix basis, double[] mean) {
112 			super(basis.getRowDimension());
113 			this.basis = basis;
114 			this.mean = mean;
115 		}
116 
117 	}
118 
119 	public static PrincipalComponentAnalysis loadPCA(File f) throws IOException {
120 		final MatFileReader reader = new MatFileReader(f);
121 		final MLSingle mean = (MLSingle) reader.getContent().get("mu");
122 		final MLSingle eigvec = (MLSingle) reader.getContent().get("proj");
123 		final Matrix basis = new Matrix(eigvec.getM(), eigvec.getN());
124 		final double[] meand = new double[eigvec.getN()];
125 		for (int j = 0; j < eigvec.getN(); j++) {
126 			// meand[i] = mean.get(i,0); ignore the means
127 			meand[j] = 0;
128 			for (int i = 0; i < eigvec.getM(); i++) {
129 				basis.set(i, j, eigvec.get(i, j));
130 			}
131 		}
132 		final PrincipalComponentAnalysis ret = new LoadedPCA(basis.transpose(), meand);
133 		return ret;
134 	}
135 
136 	public static MixtureOfGaussians loadMoG(File f) throws IOException {
137 		final MatFileReader reader = new MatFileReader(f);
138 		final MLStructure codebook = (MLStructure) reader.getContent().get("codebook");
139 
140 		final MLSingle mean = (MLSingle) codebook.getField("mean");
141 		final MLSingle variance = (MLSingle) codebook.getField("variance");
142 		final MLSingle coef = (MLSingle) codebook.getField("coef");
143 
144 		final int n_gaussians = mean.getN();
145 		final int n_dims = mean.getM();
146 
147 		final MultivariateGaussian[] ret = new MultivariateGaussian[n_gaussians];
148 		final double[] weights = new double[n_gaussians];
149 		for (int i = 0; i < n_gaussians; i++) {
150 			weights[i] = coef.get(i, 0);
151 			final DiagonalMultivariateGaussian d = new DiagonalMultivariateGaussian(n_dims);
152 			for (int j = 0; j < n_dims; j++) {
153 				d.mean.set(0, j, mean.get(j, i));
154 				d.variance[j] = variance.get(j, i);
155 			}
156 			ret[i] = d;
157 		}
158 
159 		return new MixtureOfGaussians(ret, weights);
160 	}
161 
162 }