Added --shuffle and --split options to imglab. Also added more outputs to --stats

tools/imglab/src/main.cpp

@@ -6,6 +6,7 @@
 #include "convert_pascal_v1.h"
 #include "convert_idl.h"
 #include <dlib/cmd_line_parser.h>
+#include <dlib/svm.h>
 
 #include <iostream>
 #include <fstream>
@@ -15,7 +16,7 @@
 #include <dlib/dir_nav.h>
 
 
-const char* VERSION = "0.5";
+const char* VERSION = "0.6";
 
 
 
@@ -71,6 +72,71 @@ void create_new_dataset (
 
 // ----------------------------------------------------------------------------------------
 
+int split_dataset (
+    const command_line_parser& parser
+)
+{
+    if (parser.number_of_arguments() != 1)
+    {
+        cerr << "The --split option requires you to give one XML file on the command line." << endl;
+        return EXIT_FAILURE;
+    }
+
+    const std::string label = parser.option("split").argument();
+
+    dlib::image_dataset_metadata::dataset data, data_with, data_without;
+    load_image_dataset_metadata(data, parser[0]);
+
+    data_with.name = data.name;
+    data_with.comment = data.comment;
+    data_without.name = data.name;
+    data_without.comment = data.comment;
+
+    for (unsigned long i = 0; i < data.images.size(); ++i)
+    {
+        dlib::image_dataset_metadata::image temp = data.images[i];
+
+        bool has_the_label = false;
+        // check for the label we are looking for
+        for (unsigned long j = 0; j < temp.boxes.size(); ++j)
+        {
+            if (temp.boxes[j].label == label)
+            {
+                has_the_label = true;
+                break;
+            }
+        }
+
+        if (has_the_label)
+        {
+            std::vector<dlib::image_dataset_metadata::box> boxes;
+            // remove other labels
+            for (unsigned long j = 0; j < temp.boxes.size(); ++j)
+            {
+                if (temp.boxes[j].label == label)
+                {
+                    // put only the boxes with the label we want into boxes
+                    boxes.push_back(temp.boxes[j]);
+                }
+            }
+            temp.boxes = boxes;
+            data_with.images.push_back(temp);
+        }
+        else
+        {
+            data_without.images.push_back(temp);
+        }
+    }
+
+
+    save_image_dataset_metadata(data_with, left_substr(parser[0],".") + "_with_"+label + ".xml");
+    save_image_dataset_metadata(data_without, left_substr(parser[0],".") + "_without_"+label + ".xml");
+
+    return EXIT_SUCCESS;
+}
+
+// ----------------------------------------------------------------------------------------
+
 void print_all_labels (
     const dlib::image_dataset_metadata::dataset& data
 )
@@ -100,19 +166,27 @@ void print_all_label_stats (
 )
 {
     std::map<std::string, running_stats<double> > area_stats, aspect_ratio;
+    std::map<std::string, int> image_hits;
     std::set<std::string> labels;
     for (unsigned long i = 0; i < data.images.size(); ++i)
     {
+        std::set<std::string> temp;
         for (unsigned long j = 0; j < data.images[i].boxes.size(); ++j)
         {
             labels.insert(data.images[i].boxes[j].label);
+            temp.insert(data.images[i].boxes[j].label);
 
             area_stats[data.images[i].boxes[j].label].add(data.images[i].boxes[j].rect.area());
             aspect_ratio[data.images[i].boxes[j].label].add(data.images[i].boxes[j].rect.width()/
                                                     (double)data.images[i].boxes[j].rect.height());
         }
+
+        // count the number of images for each label
+        for (std::set<std::string>::iterator i = temp.begin(); i != temp.end(); ++i)
+            image_hits[*i] += 1;
     }
 
+    cout << "Number of images: "<< data.images.size() << endl;
     cout << "Number of different labels: "<< labels.size() << endl << endl;
 
     for (std::set<std::string>::iterator i = labels.begin(); i != labels.end(); ++i)
@@ -120,6 +194,7 @@ void print_all_label_stats (
         if (i->size() != 0)
         {
             cout << "Label: "<< *i << endl;
+            cout << "   number of images:      " << image_hits[*i] << endl;
             cout << "   number of occurrences: " << area_stats[*i].current_n() << endl;
             cout << "   min box area:    " << area_stats[*i].min() << endl;
             cout << "   max box area:    " << area_stats[*i].max() << endl;
@@ -176,13 +251,20 @@ int main(int argc, char** argv)
         parser.add_option("parts","The display will allow image parts to be labeled.  The set of allowable parts "
                           "defined in a space separated list contained in <arg>.",1);
         parser.add_option("rmdiff","Remove boxes marked as difficult.");
+        parser.add_option("shuffle","Randomly shuffle the order of the images listed in file <arg>.");
+        parser.add_option("seed", "When using --shuffle, set the random seed to the string <arg>.",1);
+        parser.add_option("split", "Split the contents of an XML file into two separate files.  One containing the "
+            "images with objects labeled <arg> and another file with all the other images.  Additionally, the file "
+            "containing the <arg> labeled objects will not contain any other labels other than <arg>. "
+            "That is, the images in the first file are stripped of all labels other than the <arg> labels.",1);
 
         parser.parse(argc, argv);
 
-        const char* singles[] = {"h","c","r","l","convert","parts","rmdiff"};
+        const char* singles[] = {"h","c","r","l","convert","parts","rmdiff","seed", "shuffle", "split"};
         parser.check_one_time_options(singles);
         const char* c_sub_ops[] = {"r", "convert"};
         parser.check_sub_options("c", c_sub_ops);
+        parser.check_sub_option("shuffle", "seed");
         parser.check_incompatible_options("c", "l");
         parser.check_incompatible_options("c", "rmdiff");
         parser.check_incompatible_options("c", "rename");
@@ -269,6 +351,29 @@ int main(int argc, char** argv)
             return EXIT_SUCCESS;
         }
 
+        if (parser.option("split"))
+        {
+            return split_dataset(parser);
+        }
+
+        if (parser.option("shuffle"))
+        {
+            if (parser.number_of_arguments() != 1)
+            {
+                cerr << "The -shuffle option requires you to give one XML file on the command line." << endl;
+                return EXIT_FAILURE;
+            }
+
+            dlib::image_dataset_metadata::dataset data;
+            load_image_dataset_metadata(data, parser[0]);
+            const string default_seed = cast_to_string(time(0));
+            const string seed = get_option(parser, "seed", default_seed);
+            dlib::rand rnd(seed);
+            randomize_samples(data.images, rnd);
+            save_image_dataset_metadata(data, parser[0]);
+            return EXIT_SUCCESS;
+        }
+
         if (parser.option("stats"))
         {
             if (parser.number_of_arguments() != 1)