2 * folded-ctf is an implementation of the folded hierarchy of
3 * classifiers for object detection, developed by Francois Fleuret
6 * Copyright (c) 2008 Idiap Research Institute, http://www.idiap.ch/
7 * Written by Francois Fleuret <francois.fleuret@idiap.ch>
9 * This file is part of folded-ctf.
11 * folded-ctf is free software: you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published
13 * by the Free Software Foundation, either version 3 of the License,
14 * or (at your option) any later version.
16 * folded-ctf is distributed in the hope that it will be useful, but
17 * WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with folded-ctf. If not, see <http://www.gnu.org/licenses/>.
40 void Global::init_parser(ParamParser *parser) {
41 // The nice level of the process
42 parser->add_association("niceness", "15", false);
44 // Seed to initialize the random generator
45 parser->add_association("random-seed", "0", false);
47 // Should the pictures be b&w
48 parser->add_association("pictures-for-article", "no", false);
50 // The name of the image pool to use
51 parser->add_association("pool-name", "", false);
52 // The name of the test image pool to use
53 parser->add_association("test-pool-name", "", false);
54 // From where to load or where to save the detector
55 parser->add_association("detector-name", "default.det", false);
56 // Where to put the generated files
57 parser->add_association("result-path", "/tmp/", false);
59 // What kind of loss for the boosting
60 parser->add_association("loss-type", "exponential", false);
62 // How many images to produce/process
63 parser->add_association("nb-images", "-1", false);
64 // What is the number of the feature to show in the images
65 parser->add_association("material-feature-nb", "-1", false);
67 // What is the maximum tree depth
68 parser->add_association("tree-depth-max", "1", false);
69 // What is the proportion of negative cells we actually use during training
70 parser->add_association("proportion-negative-cells-for-training", "0.025", false);
71 // How many negative samples to sub-sample for boosting every classifier
72 parser->add_association("nb-negative-samples-per-positive", "10", false);
73 // How many features we will look at for boosting optimization
74 parser->add_association("nb-features-for-boosting-optimization", "10000", false);
75 // Do we allow head-belly registration
76 parser->add_association("force-head-belly-independence", "no", false);
77 // How many weak-learners in every classifier
78 parser->add_association("nb-weak-learners-per-classifier", "100", false);
79 // How many classifiers per level
80 parser->add_association("nb-classifiers-per-level", "25", false);
82 parser->add_association("nb-levels", "2", false);
84 // Proportion of images from the pool to use for training
85 parser->add_association("proportion-for-train", "0.5", false);
86 // Proportion of images from the pool to use for validation
87 parser->add_association("proportion-for-validation", "0.25", false);
88 // Proportion of images from the pool to use for test (negative
89 // means everything else)
90 parser->add_association("proportion-for-test", "0.25", false);
91 // During training, should we write the ROC curve estimated on the
92 // validation set (which cost a bit of computation)
93 parser->add_association("write-validation-rocs", "no", false);
95 // Should we write down the PNGs for the results of the parsing
96 parser->add_association("write-parse-images", "no", false);
98 // Should we write down the PNGs for the tags
99 parser->add_association("write-tag-images", "no", false);
101 // What is the wanted true overall positive rate
102 parser->add_association("wanted-true-positive-rate", "0.75", false);
103 // How many rates to try for the sequence of tests
104 parser->add_association("nb-wanted-true-positive-rates", "10", false);
106 // What is the minimum radius of the heads to detect. This is used
107 // as the reference size.
108 parser->add_association("min-head-radius", "25", false);
109 // What is the maximum size of the heads to detect.
110 parser->add_association("max-head-radius", "200", false);
111 // How many translation cell per radius when generating the "top
112 // level" cells for an image.
113 parser->add_association("root-cell-nb-xy-per-radius", "5", false);
115 // What is the minimum size of the windows
116 parser->add_association("pi-feature-window-min-size", "0.1", false);
118 // How many scales between two powers of two for the multi-scale
120 parser->add_association("nb-scales-per-power-of-two", "5", false);
122 // Should we display a progress bar for lengthy operations
123 parser->add_association("progress-bar", "yes", false);
126 void Global::read_parser(ParamParser *parser) {
127 niceness = parser->get_association_int("niceness");
128 random_seed = parser->get_association_int("random-seed");
129 pictures_for_article = parser->get_association_bool("pictures-for-article");
131 strncpy(pool_name, parser->get_association("pool-name"), buffer_size);
132 strncpy(test_pool_name, parser->get_association("test-pool-name"), buffer_size);
133 strncpy(detector_name, parser->get_association("detector-name"), buffer_size);
134 strncpy(result_path, parser->get_association("result-path"), buffer_size);
136 char buffer[buffer_size];
137 sprintf(buffer, "%s/log", result_path);
138 log_stream = new ofstream(buffer);
140 char *l = parser->get_association("loss-type");
141 if(strcmp(l, "exponential") == 0)
142 loss_type = LOSS_EXPONENTIAL;
143 else if(strcmp(l, "hinge") == 0)
144 loss_type = LOSS_HINGE;
145 else if(strcmp(l, "logistic") == 0)
146 loss_type = LOSS_LOGISTIC;
148 cerr << "Unknown loss type." << endl;
152 nb_images = parser->get_association_int("nb-images");
153 material_feature_nb = parser->get_association_int("material-feature-nb");
154 tree_depth_max = parser->get_association_int("tree-depth-max");
155 nb_weak_learners_per_classifier = parser->get_association_int("nb-weak-learners-per-classifier");
156 nb_classifiers_per_level = parser->get_association_int("nb-classifiers-per-level");
157 nb_levels = parser->get_association_int("nb-levels");
158 proportion_negative_cells_for_training = parser->get_association_scalar("proportion-negative-cells-for-training");
159 nb_negative_samples_per_positive = parser->get_association_int("nb-negative-samples-per-positive");
160 nb_features_for_boosting_optimization = parser->get_association_int("nb-features-for-boosting-optimization");
161 force_head_belly_independence = parser->get_association_bool("force-head-belly-independence");
162 proportion_for_train = parser->get_association_scalar("proportion-for-train");
163 proportion_for_validation = parser->get_association_scalar("proportion-for-validation");
164 proportion_for_test = parser->get_association_scalar("proportion-for-test");
165 write_validation_rocs = parser->get_association_bool("write-validation-rocs");
166 write_parse_images = parser->get_association_bool("write-parse-images");
167 write_tag_images = parser->get_association_bool("write-tag-images");
168 wanted_true_positive_rate = parser->get_association_scalar("wanted-true-positive-rate");
169 nb_wanted_true_positive_rates = parser->get_association_int("nb-wanted-true-positive-rates");
171 min_head_radius = parser->get_association_scalar("min-head-radius");
172 max_head_radius = parser->get_association_scalar("max-head-radius");
173 root_cell_nb_xy_per_radius = parser->get_association_int("root-cell-nb-xy-per-radius");
175 pi_feature_window_min_size = parser->get_association_scalar("pi-feature-window-min-size");
177 nb_scales_per_power_of_two = parser->get_association_int("nb-scales-per-power-of-two");
179 bar.set_visible(parser->get_association_bool("progress-bar"));