Peter M. Groen / openbr

Commit f61b9ea59a17e96fde9684e9ea920357ca863437

Authored by Scott Klum
2 parents 919ddb61 6ebf7eb9

Merge pull request #327 from biometrics/liblinear 

Liblinear
Inline Side-by-side
Showing 4 changed files with 391 additions and 23 deletions
openbr/plugins/liblinear.cmake 0 → 100644
  View file @f61b9ea
  1 +set(BR_WITH_LIBLINEAR OFF CACHE BOOL "Build with LibLinear")
  2 +
  3 +if(${BR_WITH_LIBLINEAR})
  4 + find_package(LibLinear REQUIRED)
  5 + set(BR_THIRDPARTY_SRC ${BR_THIRDPARTY_SRC} ${LibLinear_SRC})
  6 + set(BR_THIRDPARTY_SRC ${BR_THIRDPARTY_SRC} plugins/liblinear.cpp)
  7 +endif()
... ...
openbr/plugins/liblinear.cpp 0 → 100644
  View file @f61b9ea
  1 +#include <QTemporaryFile>
  2 +#include <opencv2/core/core.hpp>
  3 +#include <opencv2/ml/ml.hpp>
  4 +
  5 +#include "openbr_internal.h"
  6 +#include "openbr/core/opencvutils.h"
  7 +
  8 +#include <linear.h>
  9 +
  10 +using namespace cv;
  11 +
  12 +namespace br
  13 +{
  14 +
  15 +static void storeModel(const model &m, QDataStream &stream)
  16 +{
  17 + // Create local file
  18 + QTemporaryFile tempFile;
  19 + tempFile.open();
  20 + tempFile.close();
  21 +
  22 + // Save MLP to local file
  23 + save_model(qPrintable(tempFile.fileName()),&m);
  24 +
  25 + // Copy local file contents to stream
  26 + tempFile.open();
  27 + QByteArray data = tempFile.readAll();
  28 + tempFile.close();
  29 + stream << data;
  30 +}
  31 +
  32 +static void loadModel(model &m, QDataStream &stream)
  33 +{
  34 + // Copy local file contents from stream
  35 + QByteArray data;
  36 + stream >> data;
  37 +
  38 + // Create local file
  39 + QTemporaryFile tempFile(QDir::tempPath()+"/model");
  40 + tempFile.open();
  41 + tempFile.write(data);
  42 + tempFile.close();
  43 +
  44 + // Load MLP from local file
  45 + m = *load_model(qPrintable(tempFile.fileName()));
  46 +}
  47 +
  48 +class Linear : public Transform
  49 +{
  50 + Q_OBJECT
  51 + Q_ENUMS(Solver)
  52 + Q_PROPERTY(Solver solver READ get_solver WRITE set_solver RESET reset_solver STORED false)
  53 + Q_PROPERTY(float C READ get_C WRITE set_C RESET reset_C STORED false)
  54 + Q_PROPERTY(QString inputVariable READ get_inputVariable WRITE set_inputVariable RESET reset_inputVariable STORED false)
  55 + Q_PROPERTY(QString outputVariable READ get_outputVariable WRITE set_outputVariable RESET reset_outputVariable STORED false)
  56 + Q_PROPERTY(bool returnDFVal READ get_returnDFVal WRITE set_returnDFVal RESET reset_returnDFVal STORED false)
  57 + Q_PROPERTY(bool overwriteMat READ get_overwriteMat WRITE set_overwriteMat RESET reset_overwriteMat STORED false)
  58 + Q_PROPERTY(bool weight READ get_weight WRITE set_weight RESET reset_weight STORED false)
  59 +
  60 +public:
  61 + enum Solver { L2R_LR = ::L2R_LR,
  62 + L2R_L2LOSS_SVC_DUAL = ::L2R_L2LOSS_SVC_DUAL,
  63 + L2R_L2LOSS_SVC = ::L2R_L2LOSS_SVC,
  64 + L2R_L1LOSS_SVC_DUAL = ::L2R_L1LOSS_SVC_DUAL,
  65 + MCSVM_CS = ::MCSVM_CS,
  66 + L1R_L2LOSS_SVC = ::L1R_L2LOSS_SVC,
  67 + L1R_LR = ::L1R_LR,
  68 + L2R_LR_DUAL = ::L2R_LR_DUAL,
  69 + L2R_L2LOSS_SVR = ::L2R_L2LOSS_SVR,
  70 + L2R_L2LOSS_SVR_DUAL = ::L2R_L2LOSS_SVR_DUAL,
  71 + L2R_L1LOSS_SVR_DUAL = ::L2R_L1LOSS_SVR_DUAL };
  72 +
  73 +private:
  74 + BR_PROPERTY(Solver, solver, L2R_L2LOSS_SVC_DUAL)
  75 + BR_PROPERTY(float, C, 1)
  76 + BR_PROPERTY(QString, inputVariable, "Label")
  77 + BR_PROPERTY(QString, outputVariable, "")
  78 + BR_PROPERTY(bool, returnDFVal, false)
  79 + BR_PROPERTY(bool, overwriteMat, true)
  80 + BR_PROPERTY(bool, weight, false)
  81 +
  82 + model m;
  83 +
  84 + void train(const TemplateList &data)
  85 + {
  86 + Mat samples = OpenCVUtils::toMat(data.data());
  87 + Mat labels = OpenCVUtils::toMat(File::get<float>(data, inputVariable));
  88 +
  89 + problem prob;
  90 + prob.n = samples.cols;
  91 + prob.l = samples.rows;
  92 + prob.bias = -1;
  93 + prob.y = new double[prob.l];
  94 +
  95 + for (int i=0; i<prob.l; i++)
  96 + prob.y[i] = labels.at<float>(i,0);
  97 +
  98 + // Allocate enough memory for l feature_nodes pointers
  99 + prob.x = new feature_node*[prob.l];
  100 + feature_node *x_space = new feature_node[(prob.n+1)*prob.l];
  101 +
  102 + int k = 0;
  103 + for (int i=0; i<prob.l; i++) {
  104 + prob.x[i] = &x_space[k];
  105 + for (int j=0; j<prob.n; j++) {
  106 + x_space[k].index = j+1;
  107 + x_space[k].value = samples.at<float>(i,j);
  108 + k++;
  109 + }
  110 + x_space[k++].index = -1;
  111 + }
  112 +
  113 + parameter param;
  114 +
  115 + // TODO: Support grid search
  116 + param.C = C;
  117 + param.p = 1;
  118 + param.eps = FLT_EPSILON;
  119 + param.solver_type = solver;
  120 +
  121 + if (weight) {
  122 + param.nr_weight = 2;
  123 + param.weight_label = new int[2];
  124 + param.weight = new double[2];
  125 + param.weight_label[0] = 0;
  126 + param.weight_label[1] = 1;
  127 + int nonZero = countNonZero(labels);
  128 + param.weight[0] = 1;
  129 + param.weight[1] = (double)(prob.l-nonZero)/nonZero;
  130 + qDebug() << param.weight[0] << param.weight[1];
  131 + } else {
  132 + param.nr_weight = 0;
  133 + param.weight_label = NULL;
  134 + param.weight = NULL;
  135 + }
  136 +
  137 + m = *train_svm(&prob, &param);
  138 +
  139 + delete[] param.weight;
  140 + delete[] param.weight_label;
  141 + delete[] prob.y;
  142 + delete[] prob.x;
  143 + delete[] x_space;
  144 + }
  145 +
  146 + void project(const Template &src, Template &dst) const
  147 + {
  148 + dst = src;
  149 +
  150 + Mat sample = src.m().reshape(1,1);
  151 + feature_node *x_space = new feature_node[sample.cols+1];
  152 +
  153 + for (int j=0; j<sample.cols; j++) {
  154 + x_space[j].index = j+1;
  155 + x_space[j].value = sample.at<float>(0,j);
  156 + }
  157 + x_space[sample.cols].index = -1;
  158 +
  159 + float prediction;
  160 + double prob_estimates[m.nr_class];
  161 +
  162 + if (solver == L2R_L2LOSS_SVR ||
  163 + solver == L2R_L1LOSS_SVR_DUAL ||
  164 + solver == L2R_L2LOSS_SVR_DUAL ||
  165 + solver == L2R_L2LOSS_SVC_DUAL ||
  166 + solver == L2R_L2LOSS_SVC ||
  167 + solver == L2R_L1LOSS_SVC_DUAL ||
  168 + solver == MCSVM_CS ||
  169 + solver == L1R_L2LOSS_SVC)
  170 + {
  171 + prediction = predict_values(&m,x_space,prob_estimates);
  172 + if (returnDFVal) prediction = prob_estimates[0];
  173 + } else if (solver == L2R_LR ||
  174 + solver == L2R_LR_DUAL ||
  175 + solver == L1R_LR)
  176 + {
  177 + prediction = predict_probability(&m,x_space,prob_estimates);
  178 + if (returnDFVal) prediction = prob_estimates[0];
  179 + }
  180 +
  181 + if (overwriteMat) {
  182 + dst.m() = Mat(1, 1, CV_32F);
  183 + dst.m().at<float>(0, 0) = prediction;
  184 + } else {
  185 + dst.file.set(outputVariable,prediction);
  186 + }
  187 +
  188 + delete[] x_space;
  189 + }
  190 +
  191 + void store(QDataStream &stream) const
  192 + {
  193 + storeModel(m,stream);
  194 + }
  195 +
  196 + void load(QDataStream &stream)
  197 + {
  198 + loadModel(m,stream);
  199 + }
  200 +};
  201 +
  202 +BR_REGISTER(Transform, Linear)
  203 +
  204 +} // namespace br
  205 +
  206 +#include "liblinear.moc"
... ...
openbr/plugins/tree.cpp
  View file @f61b9ea
... ... @@ -35,7 +35,7 @@ static void loadModel(CvStatModel &amp;model, QDataStream &amp;stream)
35 35 stream >> data;
36 36  
37 37 // Create local file
38   - QTemporaryFile tempFile(QDir::tempPath()+"/model");
  38 + QTemporaryFile tempFile(QDir::tempPath()+"/"+QString::number(rand()));
39 39 tempFile.open();
40 40 tempFile.write(data);
41 41 tempFile.close();
... ... @@ -53,6 +53,50 @@ static void loadModel(CvStatModel &amp;model, QDataStream &amp;stream)
53 53 class ForestTransform : public Transform
54 54 {
55 55 Q_OBJECT
  56 +
  57 + void train(const TemplateList &data)
  58 + {
  59 + trainForest(data);
  60 + }
  61 +
  62 + void project(const Template &src, Template &dst) const
  63 + {
  64 + dst = src;
  65 +
  66 + float response;
  67 + if (classification && returnConfidence) {
  68 + // Fuzzy class label
  69 + response = forest.predict_prob(src.m().reshape(1,1));
  70 + } else {
  71 + response = forest.predict(src.m().reshape(1,1));
  72 + }
  73 +
  74 + if (overwriteMat) {
  75 + dst.m() = Mat(1, 1, CV_32F);
  76 + dst.m().at<float>(0, 0) = response;
  77 + } else {
  78 + dst.file.set(outputVariable, response);
  79 + }
  80 + }
  81 +
  82 + void load(QDataStream &stream)
  83 + {
  84 + loadModel(forest,stream);
  85 + }
  86 +
  87 + void store(QDataStream &stream) const
  88 + {
  89 + storeModel(forest,stream);
  90 + }
  91 +
  92 + void init()
  93 + {
  94 + if (outputVariable.isEmpty())
  95 + outputVariable = inputVariable;
  96 + }
  97 +
  98 +protected:
  99 + Q_ENUMS(TerminationCriteria)
56 100 Q_PROPERTY(bool classification READ get_classification WRITE set_classification RESET reset_classification STORED false)
57 101 Q_PROPERTY(float splitPercentage READ get_splitPercentage WRITE set_splitPercentage RESET reset_splitPercentage STORED false)
58 102 Q_PROPERTY(int maxDepth READ get_maxDepth WRITE set_maxDepth RESET reset_maxDepth STORED false)
... ... @@ -62,6 +106,15 @@ class ForestTransform : public Transform
62 106 Q_PROPERTY(bool overwriteMat READ get_overwriteMat WRITE set_overwriteMat RESET reset_overwriteMat STORED false)
63 107 Q_PROPERTY(QString inputVariable READ get_inputVariable WRITE set_inputVariable RESET reset_inputVariable STORED false)
64 108 Q_PROPERTY(QString outputVariable READ get_outputVariable WRITE set_outputVariable RESET reset_outputVariable STORED false)
  109 + Q_PROPERTY(bool weight READ get_weight WRITE set_weight RESET reset_weight STORED false)
  110 + Q_PROPERTY(TerminationCriteria termCrit READ get_termCrit WRITE set_termCrit RESET reset_termCrit STORED false)
  111 +
  112 +public:
  113 + enum TerminationCriteria { Iter = CV_TERMCRIT_ITER,
  114 + EPS = CV_TERMCRIT_EPS,
  115 + Both = CV_TERMCRIT_EPS | CV_TERMCRIT_ITER};
  116 +
  117 +protected:
65 118 BR_PROPERTY(bool, classification, true)
66 119 BR_PROPERTY(float, splitPercentage, .01)
67 120 BR_PROPERTY(int, maxDepth, std::numeric_limits<int>::max())
... ... @@ -71,10 +124,12 @@ class ForestTransform : public Transform
71 124 BR_PROPERTY(bool, overwriteMat, true)
72 125 BR_PROPERTY(QString, inputVariable, "Label")
73 126 BR_PROPERTY(QString, outputVariable, "")
  127 + BR_PROPERTY(bool, weight, false)
  128 + BR_PROPERTY(TerminationCriteria, termCrit, Iter)
74 129  
75 130 CvRTrees forest;
76 131  
77   - void train(const TemplateList &data)
  132 + void trainForest(const TemplateList &data)
78 133 {
79 134 Mat samples = OpenCVUtils::toMat(data.data());
80 135 Mat labels = OpenCVUtils::toMat(File::get<float>(data, inputVariable));
... ... @@ -88,6 +143,14 @@ class ForestTransform : public Transform
88 143 types.at<char>(samples.cols, 0) = CV_VAR_NUMERICAL;
89 144 }
90 145  
  146 + bool usePrior = classification && weight;
  147 + float priors[2];
  148 + if (usePrior) {
  149 + int nonZero = countNonZero(labels);
  150 + priors[0] = 1;
  151 + priors[1] = (float)(samples.rows-nonZero)/nonZero;
  152 + }
  153 +
91 154 int minSamplesForSplit = data.size()*splitPercentage;
92 155 forest.train( samples, CV_ROW_SAMPLE, labels, Mat(), Mat(), types, Mat(),
93 156 CvRTParams(maxDepth,
... ... @@ -95,54 +158,133 @@ class ForestTransform : public Transform
95 158 0,
96 159 false,
97 160 2,
98   - 0, // priors
  161 + usePrior ? priors : 0,
99 162 false,
100 163 0,
101 164 maxTrees,
102 165 forestAccuracy,
103   - CV_TERMCRIT_ITER | CV_TERMCRIT_EPS));
  166 + termCrit));
  167 +
  168 + if (Globals->verbose) {
  169 + qDebug() << "Number of trees:" << forest.get_tree_count();
  170 +
  171 + if (classification) {
  172 + QTime timer;
  173 + timer.start();
  174 + int correctClassification = 0;
  175 + float regressionError = 0;
  176 + for (int i=0; i<samples.rows; i++) {
  177 + float prediction = forest.predict_prob(samples.row(i));
  178 + int label = forest.predict(samples.row(i));
  179 + if (label == labels.at<float>(i,0)) {
  180 + correctClassification++;
  181 + }
  182 + regressionError += fabs(prediction-labels.at<float>(i,0));
  183 + }
  184 +
  185 + qDebug("Time to classify %d samples: %d ms\n \
  186 + Classification Accuracy: %f\n \
  187 + MAE: %f\n \
  188 + Sample dimensionality: %d",
  189 + samples.rows,timer.elapsed(),(float)correctClassification/samples.rows,regressionError/samples.rows,samples.cols);
  190 + }
  191 + }
  192 + }
  193 +};
  194 +
  195 +BR_REGISTER(Transform, ForestTransform)
  196 +
  197 +/*!
  198 + * \ingroup transforms
  199 + * \brief Wraps OpenCV's random trees framework to induce features
  200 + * \author Scott Klum \cite sklum
  201 + * \brief https://lirias.kuleuven.be/bitstream/123456789/316661/1/icdm11-camready.pdf
  202 + */
  203 +class ForestInductionTransform : public ForestTransform
  204 +{
  205 + Q_OBJECT
  206 + Q_PROPERTY(bool useRegressionValue READ get_useRegressionValue WRITE set_useRegressionValue RESET reset_useRegressionValue STORED false)
  207 + BR_PROPERTY(bool, useRegressionValue, false)
  208 +
  209 + int totalSize;
  210 + QList< QList<const CvDTreeNode*> > nodes;
  211 +
  212 + void fillNodes()
  213 + {
  214 + for (int i=0; i<forest.get_tree_count(); i++) {
  215 + nodes.append(QList<const CvDTreeNode*>());
  216 + const CvDTreeNode* node = forest.get_tree(i)->get_root();
  217 +
  218 + // traverse the tree and save all the nodes in depth-first order
  219 + for(;;)
  220 + {
  221 + CvDTreeNode* parent;
  222 + for(;;)
  223 + {
  224 + if( !node->left )
  225 + break;
  226 + node = node->left;
  227 + }
  228 +
  229 + nodes.last().append(node);
  230 +
  231 + for( parent = node->parent; parent && parent->right == node;
  232 + node = parent, parent = parent->parent )
  233 + ;
  234 +
  235 + if( !parent )
  236 + break;
  237 +
  238 + node = parent->right;
  239 + }
  240 +
  241 + totalSize += nodes.last().size();
  242 + }
  243 + }
104 244  
105   - qDebug() << "Number of trees:" << forest.get_tree_count();
  245 + void train(const TemplateList &data)
  246 + {
  247 + trainForest(data);
  248 + if (!useRegressionValue) fillNodes();
106 249 }
107 250  
108 251 void project(const Template &src, Template &dst) const
109 252 {
110 253 dst = src;
111 254  
112   - float response;
113   - if (classification && returnConfidence) {
114   - // Fuzzy class label
115   - response = forest.predict_prob(src.m().reshape(1,1));
116   - } else {
117   - response = forest.predict(src.m().reshape(1,1));
118   - }
  255 + Mat responses;
119 256  
120   - if (overwriteMat) {
121   - dst.m() = Mat(1, 1, CV_32F);
122   - dst.m().at<float>(0, 0) = response;
  257 + if (useRegressionValue) {
  258 + responses = Mat::zeros(forest.get_tree_count(),1,CV_32F);
  259 + for (int i=0; i<forest.get_tree_count(); i++) {
  260 + responses.at<float>(i,0) = forest.get_tree(i)->predict(src.m().reshape(1,1))->value;
  261 + }
123 262 } else {
124   - dst.file.set(outputVariable, response);
  263 + responses = Mat::zeros(totalSize,1,CV_32F);
  264 + int offset = 0;
  265 + for (int i=0; i<nodes.size(); i++) {
  266 + int index = nodes[i].indexOf(forest.get_tree(i)->predict(src.m().reshape(1,1)));
  267 + responses.at<float>(offset+index,0) = 1;
  268 + offset += nodes[i].size();
  269 + }
125 270 }
  271 +
  272 + dst.m() = responses;
126 273 }
127 274  
128 275 void load(QDataStream &stream)
129 276 {
130 277 loadModel(forest,stream);
  278 + if (!useRegressionValue) fillNodes();
131 279 }
132 280  
133 281 void store(QDataStream &stream) const
134 282 {
135 283 storeModel(forest,stream);
136 284 }
137   -
138   - void init()
139   - {
140   - if (outputVariable.isEmpty())
141   - outputVariable = inputVariable;
142   - }
143 285 };
144 286  
145   -BR_REGISTER(Transform, ForestTransform)
  287 +BR_REGISTER(Transform, ForestInductionTransform)
146 288  
147 289 /*!
148 290 * \ingroup transforms
... ...
share/openbr/cmake/FindLibLinear.cmake 0 → 100644
  View file @f61b9ea
  1 +find_path(LibLinear_DIR linear.h ${CMAKE_SOURCE_DIR}/3rdparty/*)
  2 +
  3 +message(${LibLinear_DIR})
  4 +mark_as_advanced(LibLinear_DIR)
  5 +include_directories(${LibLinear_DIR})
  6 +include_directories(${LibLinear_DIR}/blas)
  7 +
  8 +set(LibLinear_SRC ${LibLinear_DIR}/linear.cpp
  9 + ${LibLinear_DIR}/tron.cpp
  10 + ${LibLinear_DIR}/blas/daxpy.c
  11 + ${LibLinear_DIR}/blas/ddot.c
  12 + ${LibLinear_DIR}/blas/dnrm2.c
  13 + ${LibLinear_DIR}/blas/dscal.c)
... ...