Merge branch 'dev' of gitlab.rezometz.org:guillaume/miniprojet into dev

README.md

@@ -0,0 +1,9 @@
+# Miniprojet: Extraction des Ordres à Partir de Gestes
+
+## TODO
+
+- [ ] sauvegarder et lire des fichiers qui contiennent des descripteurs;
+- [ ] sauvegarder les poids du réseau depuis matlab et les lire en c++;
+- [ ] sauvegarder les prototypes de kmeans et les lire en c++;
+- [ ] implémenter les k-means;
+- [ ] implémenter l'algo de réseau de néuronnes;

ROS/gesture_based_control/CMakeLists.txt

@@ -1,206 +1,30 @@
 cmake_minimum_required(VERSION 2.8.3)
 project(gesture_based_control)
 
-## Compile as C++11, supported in ROS Kinetic and newer
-# add_compile_options(-std=c++11)
+add_compile_options(-std=c++11)
 
-## Find catkin macros and libraries
-## if COMPONENTS list like find_package(catkin REQUIRED COMPONENTS xyz)
-## is used, also find other catkin packages
 find_package(catkin REQUIRED COMPONENTS
   geometry_msgs
   roscpp
   std_msgs
+  cv_bridge
+  image_transport
+  dynamic_reconfigure
 )
 
-## System dependencies are found with CMake's conventions
-# find_package(Boost REQUIRED COMPONENTS system)
-
-
-## Uncomment this if the package has a setup.py. This macro ensures
-## modules and global scripts declared therein get installed
-## See http://ros.org/doc/api/catkin/html/user_guide/setup_dot_py.html
-# catkin_python_setup()
-
-################################################
-## Declare ROS messages, services and actions ##
-################################################
-
-## To declare and build messages, services or actions from within this
-## package, follow these steps:
-## * Let MSG_DEP_SET be the set of packages whose message types you use in
-##   your messages/services/actions (e.g. std_msgs, actionlib_msgs, ...).
-## * In the file package.xml:
-##   * add a build_depend tag for "message_generation"
-##   * add a build_depend and a exec_depend tag for each package in MSG_DEP_SET
-##   * If MSG_DEP_SET isn't empty the following dependency has been pulled in
-##     but can be declared for certainty nonetheless:
-##     * add a exec_depend tag for "message_runtime"
-## * In this file (CMakeLists.txt):
-##   * add "message_generation" and every package in MSG_DEP_SET to
-##     find_package(catkin REQUIRED COMPONENTS ...)
-##   * add "message_runtime" and every package in MSG_DEP_SET to
-##     catkin_package(CATKIN_DEPENDS ...)
-##   * uncomment the add_*_files sections below as needed
-##     and list every .msg/.srv/.action file to be processed
-##   * uncomment the generate_messages entry below
-##   * add every package in MSG_DEP_SET to generate_messages(DEPENDENCIES ...)
-
-## Generate messages in the 'msg' folder
-# add_message_files(
-#   FILES
-#   Message1.msg
-#   Message2.msg
-# )
-
-## Generate services in the 'srv' folder
-# add_service_files(
-#   FILES
-#   Service1.srv
-#   Service2.srv
-# )
-
-## Generate actions in the 'action' folder
-# add_action_files(
-#   FILES
-#   Action1.action
-#   Action2.action
-# )
-
-## Generate added messages and services with any dependencies listed here
-# generate_messages(
-#   DEPENDENCIES
-#   geometry_msgs#   std_msgs
-# )
-
-################################################
-## Declare ROS dynamic reconfigure parameters ##
-################################################
-
-## To declare and build dynamic reconfigure parameters within this
-## package, follow these steps:
-## * In the file package.xml:
-##   * add a build_depend and a exec_depend tag for "dynamic_reconfigure"
-## * In this file (CMakeLists.txt):
-##   * add "dynamic_reconfigure" to
-##     find_package(catkin REQUIRED COMPONENTS ...)
-##   * uncomment the "generate_dynamic_reconfigure_options" section below
-##     and list every .cfg file to be processed
-
-## Generate dynamic reconfigure parameters in the 'cfg' folder
-# generate_dynamic_reconfigure_options(
-#   cfg/DynReconf1.cfg
-#   cfg/DynReconf2.cfg
-# )
-
-###################################
-## catkin specific configuration ##
-###################################
-## The catkin_package macro generates cmake config files for your package
-## Declare things to be passed to dependent projects
-## INCLUDE_DIRS: uncomment this if your package contains header files
-## LIBRARIES: libraries you create in this project that dependent projects also need
-## CATKIN_DEPENDS: catkin_packages dependent projects also need
-## DEPENDS: system dependencies of this project that dependent projects also need
-catkin_package(
-#  INCLUDE_DIRS include
-#  LIBRARIES gesture_based_control
-#  CATKIN_DEPENDS geometry_msgs roscpp std_msgs
-#  DEPENDS system_lib
+generate_dynamic_reconfigure_options(
+  cfg/Descriptor.cfg
 )
 
-###########
-## Build ##
-###########
+catkin_package()
 
-## Specify additional locations of header files
-## Your package locations should be listed before other locations
 include_directories(
-# include
   ${catkin_INCLUDE_DIRS}
 )
 
-## Declare a C++ library
-# add_library(${PROJECT_NAME}
-#   src/${PROJECT_NAME}/gesture_based_control.cpp
-# )
+add_executable(controller src/controller.cpp)
+target_link_libraries(controller ${catkin_LIBRARIES})
 
-## Add cmake target dependencies of the library
-## as an example, code may need to be generated before libraries
-## either from message generation or dynamic reconfigure
-# add_dependencies(${PROJECT_NAME} ${${PROJECT_NAME}_EXPORTED_TARGETS} ${catkin_EXPORTED_TARGETS})
-
-## Declare a C++ executable
-## With catkin_make all packages are built within a single CMake context
-## The recommended prefix ensures that target names across packages don't collide
-# add_executable(${PROJECT_NAME}_node src/gesture_based_control_node.cpp)
-
-## Rename C++ executable without prefix
-## The above recommended prefix causes long target names, the following renames the
-## target back to the shorter version for ease of user use
-## e.g. "rosrun someones_pkg node" instead of "rosrun someones_pkg someones_pkg_node"
-# set_target_properties(${PROJECT_NAME}_node PROPERTIES OUTPUT_NAME node PREFIX "")
-
-## Add cmake target dependencies of the executable
-## same as for the library above
-# add_dependencies(${PROJECT_NAME}_node ${${PROJECT_NAME}_EXPORTED_TARGETS} ${catkin_EXPORTED_TARGETS})
-
-## Specify libraries to link a library or executable target against
-# target_link_libraries(${PROJECT_NAME}_node
-#   ${catkin_LIBRARIES}
-# )
-
-#############
-## Install ##
-#############
-
-# all install targets should use catkin DESTINATION variables
-# See http://ros.org/doc/api/catkin/html/adv_user_guide/variables.html
-
-## Mark executable scripts (Python etc.) for installation
-## in contrast to setup.py, you can choose the destination
-# install(PROGRAMS
-#   scripts/my_python_script
-#   DESTINATION ${CATKIN_PACKAGE_BIN_DESTINATION}
-# )
-
-## Mark executables for installation
-## See http://docs.ros.org/melodic/api/catkin/html/howto/format1/building_executables.html
-# install(TARGETS ${PROJECT_NAME}_node
-#   RUNTIME DESTINATION ${CATKIN_PACKAGE_BIN_DESTINATION}
-# )
-
-## Mark libraries for installation
-## See http://docs.ros.org/melodic/api/catkin/html/howto/format1/building_libraries.html
-# install(TARGETS ${PROJECT_NAME}
-#   ARCHIVE DESTINATION ${CATKIN_PACKAGE_LIB_DESTINATION}
-#   LIBRARY DESTINATION ${CATKIN_PACKAGE_LIB_DESTINATION}
-#   RUNTIME DESTINATION ${CATKIN_GLOBAL_BIN_DESTINATION}
-# )
-
-## Mark cpp header files for installation
-# install(DIRECTORY include/${PROJECT_NAME}/
-#   DESTINATION ${CATKIN_PACKAGE_INCLUDE_DESTINATION}
-#   FILES_MATCHING PATTERN "*.h"
-#   PATTERN ".svn" EXCLUDE
-# )
-
-## Mark other files for installation (e.g. launch and bag files, etc.)
-# install(FILES
-#   # myfile1
-#   # myfile2
-#   DESTINATION ${CATKIN_PACKAGE_SHARE_DESTINATION}
-# )
-
-#############
-## Testing ##
-#############
-
-## Add gtest based cpp test target and link libraries
-# catkin_add_gtest(${PROJECT_NAME}-test test/test_gesture_based_control.cpp)
-# if(TARGET ${PROJECT_NAME}-test)
-#   target_link_libraries(${PROJECT_NAME}-test ${PROJECT_NAME})
-# endif()
-
-## Add folders to be run by python nosetests
-# catkin_add_nosetests(test)
+add_executable(descripteur src/descripteur.cpp)
+target_link_libraries(descripteur ${catkin_LIBRARIES})
+#add_dependencies(descripteur ${PROJECT_NAME}_gencfg)

ROS/gesture_based_control/cfg/Descriptor.cfg

@@ -0,0 +1,11 @@
+#!/usr/bin/env python
+PACKAGE = "gesture_based_control"
+
+from dynamic_reconfigure.parameter_generator_catkin import *
+
+gen = ParameterGenerator()
+
+gen.add("cmax", int_t, 0, "cmax",      50,  0, 100)
+gen.add("threshold", int_t, 0, "threshold", 50,  0, 255)
+
+exit(gen.generate(PACKAGE, "gesture_based_control", "Descriptors"))

ROS/gesture_based_control/launch/gesture_control.launch

@@ -0,0 +1,14 @@
+<launch>
+  <node name="usb_cam" pkg="usb_cam" type="usb_cam_node">
+    <param name="pixel_format" value="yuyv"/>
+  </node>
+  <node name="rqt_image_view" pkg="rqt_image_view" type="rqt_image_view"/>
+  <node name="controller" pkg="gesture_based_control" type="controller" output="screen">
+    <remap from="/cmd_vel" to="/turtle1/cmd_vel"/>
+  </node>
+  <node name="turtle" pkg="turtlesim" type="turtlesim_node"/>
+  <node name="descripteur" pkg="gesture_based_control" type="descripteur" output="screen">
+    <remap from="/image_raw" to="/usb_cam/image_raw"/>
+  </node>
+  <node name="rqt_reconfigure" pkg="rqt_reconfigure" type="rqt_reconfigure"/>
+</launch>

ROS/gesture_based_control/src/controller.cpp

@@ -0,0 +1,38 @@
+#include <ros/ros.h>
+#include <std_msgs/String.h>
+#include <geometry_msgs/Twist.h>
+
+void callback(const std_msgs::String& msg, geometry_msgs::Twist& cmd) {
+  if (msg.data == "avance") {
+    cmd.linear.x = 1.0;
+    cmd.angular.z = 0.0;
+  } else if (msg.data == "arret") {
+    cmd = geometry_msgs::Twist();
+  } else if (msg.data == "gauche") {
+    cmd.angular.z = 1.0;
+    cmd.linear.x = 0.0;
+  } else if (msg.data == "droite") {
+    cmd.angular.z = -1.0;
+    cmd.linear.x = 0.0;
+  }
+}
+
+int main(int argc, char** argv) {
+  ros::init(argc, argv, "controller");
+
+  ros::NodeHandle n;
+  ros::Publisher pub = n.advertise<geometry_msgs::Twist>("/cmd_vel", 0);
+
+  geometry_msgs::Twist cmd;
+  ros::Rate rate(10);
+
+  boost::function<void(const std_msgs::String&)> cb = boost::bind(callback, _1, boost::ref(cmd));
+  ros::Subscriber sub = n.subscribe<std_msgs::String&>("/order", 0, cb);
+
+  while(ros::ok()) {
+    pub.publish(cmd);
+    ros::spinOnce();
+    rate.sleep();
+  }
+  return 0;
+}

ROS/gesture_based_control/src/descripteur.cpp

@@ -0,0 +1,69 @@
+#include <ros/ros.h>
+#include <opencv2/opencv.hpp>
+#include <image_transport/image_transport.h>
+#include <cv_bridge/cv_bridge.h>
+#include <std_msgs/String.h>
+#include <dynamic_reconfigure/server.h>
+
+#include <gesture_based_control/DescriptorsConfig.h>
+#include "math.hpp"
+#include "file.hpp"
+#include "knn.hpp"
+
+void callback(gesture_based_control::DescriptorsConfig &config, uint32_t level, int& cmax, int& threshold) {
+  threshold = config.threshold;
+}
+
+void callback(const sensor_msgs::ImageConstPtr& msg, int& threshold, int& cmax, ros::Publisher& order_pub, image_transport::Publisher& image_pub, math::dataset& dataset) {
+  cv_bridge::CvImagePtr cv_ptr;
+  std_msgs::String s;
+
+  try {
+    cv_ptr = cv_bridge::toCvCopy(msg, sensor_msgs::image_encodings::RGB8);
+  }catch(cv_bridge::Exception& e) {
+    ROS_ERROR("cv_bridge exception: %s", e.what());
+    return; 
+  }
+
+  cv::Mat& input = cv_ptr->image;
+  cv::Mat binaire(input.rows, input.cols, CV_8UC1);
+  math::filter(input, binaire, threshold);
+  cv::GaussianBlur(input, input, cv::Size(7,7), 1.5, 1.5);
+  
+  std::vector<math::contour> contours;
+  std::vector<cv::Vec4i> hierarchy;
+  cv::findContours(binaire, contours, hierarchy, CV_RETR_EXTERNAL, CV_CHAIN_APPROX_NONE);
+  if (contours.size() != 0) {
+    int index = math::max_cont(contours);
+    math::csignal desc = math::descriptors(contours[index], cmax);
+    s.data = predict(desc, dataset, 5, 2);
+    order_pub.publish(s);
+
+    cv::drawContours(input, contours, index, 255);
+  }
+  image_pub.publish(cv_bridge::CvImage(msg->header, "rgb8", input).toImageMsg());
+}
+
+int main(int argc, char** argv) {
+  ros::init(argc, argv, "descripteur");
+
+  int threshold = 25;
+  int cmax = 10;
+
+  math::dataset dataset = load_csv("/home/guillaume/Documents/3A_supelec/miniprojet/learning/descripteurs/dataset50.csv", 21);
+
+  ros::NodeHandle n;
+  image_transport::ImageTransport it(n);
+
+  ros::Publisher order_pub = n.advertise<std_msgs::String>("/order", 100);
+  image_transport::Publisher image_pub = it.advertise("/image_out", 1);
+  image_transport::Subscriber sub = it.subscribe("/image_raw", 1, boost::bind(callback, _1, boost::ref(threshold), boost::ref(cmax), boost::ref(order_pub), boost::ref(image_pub), boost::ref(dataset)));
+
+  dynamic_reconfigure::Server<gesture_based_control::DescriptorsConfig> server;
+  dynamic_reconfigure::Server<gesture_based_control::DescriptorsConfig>::CallbackType f;
+
+  f = boost::bind(&callback, _1, _2, boost::ref(cmax), boost::ref(threshold));
+  server.setCallback(f);
+  
+  ros::spin();
+}

ROS/gesture_based_control/src/math.hpp

@@ -0,0 +1 @@
+/home/guillaume/Documents/3A_supelec/miniprojet/tests/src/math.hpp

learning/code/PMC_training.m

@@ -0,0 +1,29 @@
+function [net, rt] = PMC_training(entree, sortie, n_cache, nb_apprentissage)
+
+[n_entree, nb_echantillons] = size(entree);
+[n_sortie, ~] = size(sortie);
+
+% net = newff(entree, sortie, n_cache, {'tansig' 'tansig'}, 'trainscg');
+net = feedforwardnet(n_cache, 'trainscg');
+
+net.trainParam.epochs = 1000; % Le nombre de cycle d’apprentissage est fixé à 1000
+
+net.trainParam.lr = 0.02; % Le pas d’apprentissage est égal à 0.02
+
+net.trainParam.show = 100; % Des informations sur les performances du réseau sont affichées tous les 100 cycles d’apprentissage
+
+net.trainParam.goal = 1e-10; % L’algorithme d’apprentissage s’arrête lorsque l’erreur quadratique moyenne est inférieure à 1e-10
+
+net.trainParam.min_grad = 1e-10; % L’algorithme d’apprentissage s’arrête lorsque le module du gradient est inférieur à 1e-10
+
+net.divideParam.trainRatio = nb_apprentissage / nb_echantillons; 
+net.divideParam.valRatio = 0;  %  On n'utilise pas d'ensemble de validation.
+net.divideParam.testRatio = (nb_echantillons - nb_apprentissage) / nb_echantillons; 
+
+net = train(net, entree, sortie);
+
+entree_test = entree(:,(nb_apprentissage + 1): nb_echantillons);
+sortie_test = sortie(:,(nb_apprentissage + 1): nb_echantillons);
+rt = sim(net, entree_test, [], [], sortie_test);
+end
+

learning/code/algo.c

@@ -0,0 +1,176 @@
+#define _CRT_SECURE_NO_DEPRECATE
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <malloc.h>
+#include <math.h>
+#include "constant.h"
+#include "global.h"
+#include "algo.h"
+
+void Erreur(char * chaine,int numero)
+{
+	fprintf(stderr,chaine);
+	fprintf(stderr,"\nerreur %d\n",numero);
+	exit(numero);
+}
+
+int LireVecteurs (void)
+{
+	int k;
+	if((ent=fopen(noment,"rb"))==NULL)
+		return 1;
+	fread(&M,sizeof(int),1,ent);
+	fread(&dimvec,sizeof(int),1,ent);
+	fread(&dico,sizeof(int),1,ent);
+	fread(&numiter,sizeof(int),1,ent);
+	if((vecteurs=(float**)calloc(M,sizeof(float *)))==NULL)
+		return 3;
+	for(k=0;k<M;k++)
+	{
+		if((vecteurs[k]=(float*)calloc(dimvec,sizeof(float)))==NULL)
+			return 4;
+	}
+	for(k=0;k<M;k++)
+		fread(vecteurs[k],sizeof(float),dimvec,ent);
+	fclose(ent);
+	return 0;
+}
+
+int InitDictionnaire(void)
+{
+	int k;
+	if((vecteurs_classes=(float**)calloc(dico,sizeof(float *)))==NULL)
+		return 1;
+	for(k=0;k<dico;k++)
+	{
+		if((vecteurs_classes[k]=(float*)calloc(dimvec,sizeof(float)))==NULL)
+			return 2;
+	}
+	if((vecteurs_final=(float**)calloc(dico,sizeof(float *)))==NULL)
+		return 3;
+	for(k=0;k<dico;k++)
+	{
+		if((vecteurs_final[k]=(float*)calloc(dimvec,sizeof(float)))==NULL)
+			return 4;
+	}
+	if((index_final=(int*)calloc(M,sizeof(int)))==NULL)
+		return 5;
+	if((popul_final=(int*)calloc(dico,sizeof(int)))==NULL)
+		return 6;
+	if((index_classes=(int*)calloc(M,sizeof(int)))==NULL)
+		return 7;
+	if((popul_classes=(int*)calloc(dico,sizeof(int)))==NULL)
+		return 8;
+	return 0;
+}
+
+int CalculDictionnaire(void)
+{
+	int n,k,p,index;
+	double energie_prec,distance,distancemin;
+	char fini;
+	for(k=0;k<dico;k++)
+	{
+		index=(int)floor(((double)(M-1)*(double)(rand()))/(double)RAND_MAX);
+		for(n=0;n<dimvec;n++)
+			vecteurs_classes[k][n]=vecteurs[index][n];
+	}
+	energie=0.0;
+	etape=0;
+	fini=0;
+	while(!fini)
+	{
+		memset(popul_classes,0,dico*sizeof(int));
+		energie_prec=energie;
+		energie=0.0;
+		for(k=0;k<M;k++)
+		{
+			index=0;
+			distancemin=0.0;
+			for(n=0;n<dimvec;n++)
+			{
+				distancemin+=
+					(vecteurs[k][n]-vecteurs_classes[0][n])*
+					(vecteurs[k][n]-vecteurs_classes[0][n]);
+			}
+			for(p=1;p<dico;p++)
+			{
+				distance=0.0;
+				for(n=0;n<dimvec;n++)
+				{
+					distance+=
+						(vecteurs[k][n]-vecteurs_classes[p][n])*
+						(vecteurs[k][n]-vecteurs_classes[p][n]);
+				}
+				if(distance<distancemin)
+				{
+					distancemin=distance;
+					index=p;
+				}
+			}
+			index_classes[k]=index;
+			popul_classes[index]+=1;
+			energie+=distancemin;
+		}
+		for(p=0;p<dico;p++)
+		{
+			if(popul_classes[p])
+			{
+				memset(vecteurs_classes[p],0,dimvec*sizeof(float));
+			}
+		}
+		for(k=0;k<M;k++)
+		{
+			index=index_classes[k];
+			if(popul_classes[index])
+			{
+				for(n=0;n<dimvec;n++)
+				{
+					vecteurs_classes[index][n]+=vecteurs[k][n];
+				}
+			}
+		}
+		for(p=0;p<dico;p++)
+		{
+			if(popul_classes[p])
+			{
+				for(n=0;n<dimvec;n++)
+				{
+					vecteurs_classes[p][n]/=popul_classes[p];
+				}
+			}
+		}
+		if(etape!=0)
+		{
+			if(((energie_prec-energie)/energie)<SEUIL)
+				fini=1;
+		}
+		etape++;
+	}
+	return 0;
+}
+
+int SauveClasses(void)
+{
+	int k;
+	if((sor=fopen(nomsor,"wb"))==NULL)
+		return 1;
+	fwrite(&dimvec,sizeof(int),1,sor);
+	fwrite(&dico,sizeof(int),1,sor);
+	for(k=0;k<dico;k++)
+		fwrite(vecteurs_final[k],sizeof(float),dimvec,sor);
+	fwrite(popul_final,sizeof(int),dico,sor);
+	fclose(sor);
+	return 0;
+}
+
+int SauveCode(void)
+{
+	if((cod=fopen(nomcod,"wb"))==NULL)
+		return 1;
+	fwrite(index_final,sizeof(int),M,sor);
+	fclose(sor);
+	return 0;
+}

learning/code/algo.h

@@ -0,0 +1,6 @@
+void Erreur(char *,int);
+int LireVecteurs(void);
+int InitDictionnaire(void);
+int CalculDictionnaire(void);
+int SauveClasses(void);
+int SauveCode(void);

learning/code/constant.h

@@ -0,0 +1,10 @@
+#define NOM		50
+#define SEUIL	1e-5
+
+typedef struct
+{
+	int index;
+	int popul;
+	float * vecteur;
+}
+TRI;

learning/code/data_set.m

@@ -0,0 +1,4 @@
+function [image]=data_set(i, n)
+
+
+end

learning/code/descripteurfouriernorm.m

@@ -0,0 +1,28 @@
+function [coeff,num]=descripteurfouriernorm(z,cmax)
+cmin=-cmax;
+z_moy=mean(z);
+longc=length(z);
+% on calcule les coefficients de Fourier
+TC=fft(z-z_moy)/longc;
+num=cmin:cmax;
+% on sélectionne les coefficients entre cmin et cmax
+coeff=zeros(cmax-cmin+1,1);
+coeff(end-cmax:end)=TC(1:cmax+1);
+coeff(1:-cmin)=TC(end+cmin+1:end);
+
+% on retourne la séquence si le parcours est dans le
+% sens inverse du sens trigonométrique
+if abs(coeff(num==-1))>abs(coeff(num==1))
+    coeff=coeff(end:-1:1);
+end
+
+% corrections de phase pour normaliser
+% par rapport à la rotation et l'origine
+% du signal z
+Phi=angle(coeff(num==-1).*coeff(num==1))/2;
+coeff=coeff*exp(-1i*Phi);
+theta=angle(coeff(num==1));
+coeff=coeff.*exp(-1i*num'*theta);
+
+% correction pour normaliser la taille
+coeff=coeff/abs(coeff(num==1));

learning/code/global.c

@@ -0,0 +1,25 @@
+#define _CRT_SECURE_NO_DEPRECATE
+
+#include <stdio.h>
+#include "constant.h"
+
+char noment[NOM];
+char nomsor[NOM];
+char nomcod[NOM];
+int dimvec,M,dico,numiter;
+
+FILE * ent,* sor,* cod;
+
+float ** vecteurs;
+float ** vecteurs_classes;
+int * index_classes;
+int * popul_classes;
+float ** vecteurs_final;
+int * index_final;
+int * popul_final;
+int etape;
+double energie;
+double energieminimale;
+
+TRI * buftri;
+

learning/code/global.h

@@ -0,0 +1,25 @@
+
+
+
+
+
+extern char noment[NOM];
+extern char nomsor[NOM];
+extern char nomcod[NOM];
+extern int dimvec,M,dico,numiter;
+
+extern FILE * ent,* sor,* cod;
+
+extern float ** vecteurs;
+extern float ** vecteurs_classes;
+extern int * index_classes;
+extern int * popul_classes;
+extern float ** vecteurs_final;
+extern int * index_final;
+extern int * popul_final;
+extern int etape;
+extern double energie;
+extern double energieminimale;
+
+extern TRI * buftri;
+

learning/code/kmoyennes.m

@@ -0,0 +1,41 @@
+% kmoyennes.m
+% [vecteursliste,code,occur]=kmoyennes(vecteurs,dico,numiter)
+% vecteurs :  vecteurs de la base (rangés en colonnes)
+% dico : nombre de classes
+% numiter : nombre d'essais pour trouver le dictionnaire
+% vecteursliste : contient les vecteurs prototypes
+% code : contient les index des classes
+% occur : nombre d'éléments dans chaque classe
+
+function [vecteursliste,code,occur]=kmoyennes(vecteurs,dico,numiter)
+
+% nombre de lignes du tableau vecteurs : dimension des vecteurs
+dimvec=size(vecteurs,1);
+% nombre de colonnes du tableau vecteurs : nombre de vecteurs
+M=size(vecteurs,2);
+
+% ecriture des fichiers nécessaires au fonctionnment du programme quantvec
+fid=fopen('vecteurs','w');
+fwrite(fid,M,'int');
+fwrite(fid,dimvec,'int');
+fwrite(fid,dico,'int');
+fwrite(fid,numiter,'int');
+fwrite(fid,vecteurs,'float');
+fclose(fid);
+
+% lancement du programme quantvec
+unix('./quantvec vecteurs dict code');
+
+% lecture des fichiers résultat
+fid=fopen('dict','r');
+dimvec=fread(fid,1,'int');
+dico=fread(fid,1,'int');
+vecteursliste=fread(fid,[dimvec dico],'float');
+occur=fread(fid,dico,'int');
+fclose(fid);
+disp('vecteurs dans la liste')
+disp([dimvec dico])
+
+fid=fopen('code','r');
+code=fread(fid,M,'int');
+fclose(fid);

learning/code/learning.m

@@ -0,0 +1,92 @@
+close all
+
+%acceptable threshold values: 15-30
+threshold = 20;
+cmax = 10;
+cmin = -cmax;
+n_classes = 16;
+iterations = 50;
+N = 200;
+
+dataset = dir('../images/*/*.jpg');
+dataset_size = length(dataset);
+
+vecteurs=zeros(2*(cmax-cmin+1),dataset_size); %TODO: renommer en vectors
+
+classes = []; %colonne [avance; arret; gauche; droite; rejet]
+
+% c'est lent
+% s'assurer que l'on choisit toutes les images
+% 
+n=1;
+while n<dataset_size+1
+    % choix d'une image aleatoire
+    choix = 1 + floor(dataset_size*rand(dataset_size, 1));
+    % extraction de l'image du dataset
+    image = dataset(choix(1));
+    % lecture de l'image
+    img = imread([image.folder '/' image.name]);
+    % filtrage de la couleur de la peau
+    binary = rgb_filter(img, threshold);
+    % determination du contour
+    c = contourc(binary);
+    if size(c,2)~=0
+        
+        % Determination du contour de taille max
+        cont = max_contour(c);
+        % transformation en signal complexe
+        z = cont(:,1) + 1i*cont(:,2);
+        if length(z)>11
+            
+            % calcul des descripteurs de Fourier
+            [coeff,ncoeff]=descripteurfouriernorm(z,cmax);
+            % Extraction des composantes
+            vecteurs(:,n)=[real(coeff);imag(coeff)];
+            % Ajout de la classe correspondante
+            deb = dataset(choix(1)).name(1:2);
+            if deb == 'av'
+                classes = [classes , [1;0;0;0;0]];         
+            elseif deb == 'ar'
+               classes = [classes , [0;1;0;0;0]];       
+            elseif deb == 'ga'
+               classes = [classes , [0;0;1;0;0]];        
+            elseif deb == 'dr'
+               classes = [classes , [0;0;0;1;0]];        
+            else
+               classes = [classes , [0;0;0;0;1]];     
+            end 
+            
+            
+            % affichage de l'avancement
+            disp(n/dataset_size);
+            n = n+1;
+        end
+    end  
+    
+end
+
+% utilisation de l'algorithme des kmeans
+kmeans = kmoyennes(vecteurs, n_classes, iterations);
+
+% affichage des prototypes
+figure
+for n=1:n_classes
+    contfil=resconstrdesfour(kmeans(1:end/2,n)+1i*kmeans(end/2+1:end,n),N,cmax);
+    subplot(4,4,n)
+    h=plot(real(contfil),imag(contfil),'-',real(contfil(1)),imag(contfil(1)),'o');
+    title(['prototype ' int2str(n)])
+    set(h(1),'LineWidth',2)
+    set(h(2),'LineWidth',3)
+    grid on
+    axis equal
+    axis ij
+    drawnow
+end
+
+% utilisation de l'algorithme du perceptron multicouches
+
+[net, resultats_test] = PMC_training(vecteurs, classes, 10, 1100);
+
+%faire une prediction : y = net(x)
+
+

learning/code/makefile

@@ -0,0 +1,8 @@
+all : quantvec
+
+quantvec : *.c *.h
+	g++ -O3 -Wall *.c -o quantvec
+
+clean :
+	rm quantvec
+

learning/code/max_contour.m

@@ -0,0 +1,18 @@
+function [cont]=max_contour(contours)
+i=1;
+id=1;
+max = 0;
+while i+contours(2, id) < size(contours, 2)
+    contours(2,i);
+    if contours(2,i) > max
+        max = contours(2,i);
+        id = i;
+    end
+    i=i+1+contours(2,i);
+end
+
+cont = zeros(max, 2);
+cont(1:end,1) = contours(1, id+1:id+max);
+cont(1:end,2) = contours(2, id+1:id+max);
+
+end

learning/code/quantvec.c

@@ -0,0 +1,63 @@
+#define _CRT_SECURE_NO_DEPRECATE
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <time.h>
+#include <string.h>
+#include <malloc.h>
+#include <math.h>
+#include "constant.h"
+#include "global.h"
+#include "algo.h"
+
+int main(int ac,char ** av)
+{
+	int erreur,iter,p;
+	if(ac != 4)
+		Erreur((char*)"Il faut 3 parametres",1);
+	strncpy(noment,av[1],NOM-1);
+	strncpy(nomsor,av[2],NOM-1);
+	strncpy(nomcod,av[3],NOM-1);
+	
+	srand( (unsigned)time( NULL ) );
+	
+	if((erreur=LireVecteurs())!=0)
+		Erreur((char*)"Erreur de lecture des vecteurs",erreur);
+	fprintf(stderr,"nombre de vecteurs : %d\n",M);
+	fprintf(stderr,"dimension des vecteurs : %d\n",dimvec);
+	fprintf(stderr,"taille du dictionnaire : %d\n",dico);
+	fprintf(stderr,"nombre d'iterations : %d\n",numiter);
+	if((erreur=InitDictionnaire())!=0)
+		Erreur((char*)"Erreur d'initialisation du dictionnaire",erreur);
+
+	iter=0;
+	if((erreur=CalculDictionnaire())!=0)
+		Erreur((char*)"Erreur de calcul itératif du dictionnaire",erreur);
+	energieminimale=energie;
+	memcpy(index_final,index_classes,M*sizeof(int));
+	memcpy(popul_final,popul_classes,dico*sizeof(int));
+	for(p=0;p<dico;p++)
+		memcpy(vecteurs_final[p],vecteurs_classes[p],dimvec*sizeof(float));
+	fprintf(stderr,"iteration %d, distorsion %lf en %d etapes\n",iter,energie/((double)M*(double)dimvec),etape);
+
+	for(iter=1;iter<numiter;iter++)
+	{
+		if((erreur=CalculDictionnaire())!=0)
+			Erreur((char*)"Erreur de calcul itératif du dictionnaire",erreur);
+		if(energie<energieminimale)
+		{
+			energieminimale=energie;
+			memcpy(index_final,index_classes,M*sizeof(int));
+			memcpy(popul_final,popul_classes,dico*sizeof(int));
+			for(p=0;p<dico;p++)
+				memcpy(vecteurs_final[p],vecteurs_classes[p],dimvec*sizeof(float));
+		}
+		fprintf(stderr,"iteration %d, distorsion %lf en %d etapes\n",iter,energie/((double)M*(double)dimvec),etape);
+	}
+
+	fprintf(stderr,"distorsion finale %lf\n",energieminimale/((double)M*(double)dimvec));
+	if((erreur=SauveClasses())!=0)
+		Erreur((char*)"Erreur de sauvegarde des classes",erreur);
+	if((erreur=SauveCode())!=0)
+		Erreur((char*)"Erreur de sauvegarde du code",erreur);
+}

learning/code/resconstrdesfour.m

@@ -0,0 +1,11 @@
+function z_fil=resconstrdesfour(coeff,N,cmax)
+
+cmin=-cmax;
+TC=zeros(N,1);
+
+TC(1:cmax+1)=coeff(end-cmax:end);
+TC(end+cmin+1:end)=coeff(1:-cmin);
+
+z_fil=ifft(TC)*N;
+
+z_fil=[z_fil;z_fil(1)];

learning/code/rgb_filter.m

@@ -0,0 +1,23 @@
+function [filtered_img]=rgb_filter(img, threshold)
+height = size(img, 1);
+width = size(img, 2);
+depth = size(img, 3);
+
+filtered_img = zeros(height, width, 1);
+
+for i=1:height
+    for j=1:width
+        r = img(i, j, 1);
+        g = img(i, j, 2);
+        b = img(i, j, 3);
+        if (r > g && r > b)
+            if (r-b > threshold || r-g > threshold)
+                filtered_img(i, j, 1) = 1;
+            else
+                filtered_img(i, j, 1) = 0;
+            end
+        end
+    end
+end
+
+end

learning/code/traitement.h

@@ -0,0 +1,3 @@
+int InitInfo(void);
+void TraiteErreur(int,char*);
+DWORD WINAPI Traitement(LPVOID);