Merge pull request #2 from Dagiopia/master

all in one test and mods

CMakeLists.txt

@@ -53,7 +53,10 @@ add_executable(SaliencyTest examples/sense/vision/SaliencyTest.cpp)
 target_link_libraries(SaliencyTest sensor_vision ${OpenCV_LIBS} ${EXTRA_LIBS})   
 
 add_executable(SaliencyTest2 examples/sense/vision/SaliencyTest2.cpp)
-target_link_libraries(SaliencyTest2 sensor_vision ${OpenCV_LIBS} ${EXTRA_LIBS})   
+target_link_libraries(SaliencyTest2 sensor_vision ${OpenCV_LIBS} ${EXTRA_LIBS})  
+
+add_executable(AllTest examples/sense/vision/AllTest.cpp)
+target_link_libraries(AllTest sensor_vision ${OpenCV_LIBS} ${EXTRA_LIBS})  
 
 if(raspicam_FOUND)
 add_executable(RaspiVisionTest examples/sense/vision/RaspiVisionTest.cpp)

examples/sense/vision/AllTest.cpp

@@ -0,0 +1,163 @@
+/*
+  Project: OpenCogER
+  File: AllTest.cpp
+  Author: Dagim Sisay
+  License: AGPL
+  Date: April, 2018
+*/
+
+
+#include <iostream>
+#include <string>
+#include <signal.h>
+
+#include "sense/vision/CamCapture.hpp"
+#include "sense/vision/ITColor2Gray.hpp"
+#include "sense/vision/ITEqualizeHist.hpp"
+#include "sense/vision/ITDetectFace.hpp"
+#include "sense/vision/ITDetectSmile.hpp"
+#include "sense/vision/FacialLandMark.hpp"
+#include "sense/vision/ITDetectHand.hpp"
+#include "sense/vision/FingersCount.hpp"
+#include "sense/vision/DSaliency.hpp"
+#include "sense/vision/BoxTrackerThread.hpp"
+
+
+#define _NEED_GUI_
+#define _NEED_TIME_INFO_
+
+#define NO_FLMS 26
+#define SCALE 0.1
+
+using namespace std;
+using namespace cv;
+
+
+float avg_time_pf, en_time, acc, avg_oh, avg_nh, avg_nf, avg_of;
+uint64_t st_time;
+int n_f, nh, oh, nf, of;
+
+
+void print_report()
+{
+	avg_time_pf = acc / n_f;
+	avg_nh /= nh;
+	avg_oh /= oh;
+	avg_of /= of;
+	avg_nf /= nf;
+	printf("\nAverage Time per Frame (no Face): %f\n", avg_nf);
+	printf("Average Time per Frame (on Face): %f\n", avg_of);
+	printf("Average Time per Frame (no Hand): %f\n", avg_nh);
+	printf("Average Time per Frame (on Hand): %f\n", avg_oh);
+	printf("Average Time per Frame: %f\n\n", avg_time_pf);
+}
+
+
+void sigint_handler(int sig)
+{
+	cout<<"Caught SIGINT... Exiting!\n";
+	#ifdef _NEED_TIME_INFO_
+	print_report();
+	#endif //_NEED_TIME_INFO_
+	exit(0);
+}
+
+
+int main(int argc, char **argv)
+{
+	#ifdef _NEED_TIME_INFO_
+	avg_nh = avg_oh = avg_of = avg_nf = avg_time_pf = 0;
+	oh = nh = of = nf = n_f = 0;
+	#endif //_NEED_TIME_INFO_
+	bool face_found = false;
+	bool hand_found = false;
+
+	signal(SIGINT, sigint_handler);
+	CamCapture cc("cam", 320, 240, 0, 20);
+	if (!cc.isOk()){std::cout<<std::endl<<cc.getState()<<std::endl;return -1;}
+
+	Mat frame, img, img2;
+
+	ITColor2Gray c2g("c2g1");
+	ITEqualizeHist eh("eh1");
+	ITDetectFace fcd("fcd1");
+	ITDetectSmile smd("smd1");
+	FacialLandMark flm;
+	ITDetectHand dh("dh1");
+	FingersCount fc(true);
+	DSaliency sal_d(SAL_STATIC, SAL_FINE_GRAINED);
+
+	vector <Rect> faces, hands;
+	vector <facial_lms> f_lms;
+	facial_lms shape;
+	Rect2d box2d;
+	Rect box;
+	vector<vector<Point> > cntrs;
+	vector<Vec4i> hier;
+
+
+	BoxTrackerThread *bt;
+
+	while (true)
+	{
+		#ifdef _NEED_TIME_INFO_ 
+		st_time = getTickCount();
+		#endif //_NEED_TIME_INFO_
+		frame = cc.getCurrentFrame();
+		img = eh.Transform(c2g.Transform(frame));
+		faces = fcd.Transform(img);
+
+		f_lms.clear();
+		flm.get_lm_points(frame, faces, &f_lms); 
+		for (int i = 0; i < f_lms.size(); ++i){
+			face_found = true;
+			shape = f_lms[i];
+			#ifdef _NEED_GUI_
+			for (int j = 0; j < NO_FLMS; ++j)
+				circle(frame, Size(shape.part(j).x(), shape.part(j).y()), 1, 
+				       CV_RGB(0, 255, 0), 1.5);
+			#endif //_NEED_GUI_
+		}
+		threshold(img, img, 70, 255, CV_THRESH_BINARY_INV | CV_THRESH_OTSU);
+		hands = dh.Transform(img);
+		for (int i = 0; i < hands.size(); ++i){
+			hand_found = true;
+			img2 = Mat(img, Rect(hands[i].x - hands[i].x*SCALE,
+					    hands[i].y - hands[i].y*SCALE,
+					    hands[i].width + hands[i].width*SCALE,
+					    hands[i].height + hands[i].height*SCALE));
+			#ifdef _NEED_GUI_
+			putText(frame, to_string(fc.num_fingers(img2)), Point(20, 20),
+				FONT_HERSHEY_COMPLEX_SMALL, 1.5, CV_RGB(255, 0, 0), 2, CV_AA);
+			cvtColor(img2, img2, COLOR_GRAY2BGR);
+			for(size_t j = 0; j < fc.f_tips.size(); ++j) 
+				circle(img2, fc.f_tips[j], 2, CV_RGB(255, 0, 0), 2);
+			imshow("hand", img2);
+			#endif //_NEED_GUI_
+		}
+		#ifdef _NEED_GUI_
+		imshow("feed", frame);
+		if(27 == waitKey(10))
+			break;
+		#endif //_NEED_GUI_
+
+		#ifdef _NEED_TIME_INFO_
+		en_time = (float)((getTickCount() - st_time) / getTickFrequency());
+		if (hand_found){ avg_oh += en_time; oh++; }
+		else { avg_nh += en_time; nh++; }
+		if (face_found){ avg_of += en_time; of++; }
+		else { avg_nf += en_time; nf++; }
+		acc += en_time;
+		n_f++;
+		#endif //_NEED_TIME_INFO_
+
+		face_found = false;
+		hand_found = false;
+	} // while (true)
+
+	#ifdef _NEED_TIME_INFO_
+	print_report();
+	#endif //_NEED_TIME_INFO_
+
+return 0;
+}

examples/sense/vision/FaceLandMarkTest.cpp

@@ -70,13 +70,13 @@ int main(int argc, char** argv)
 	for (uint8_t idx = 0; idx < f_lms.size(); idx++)
 	{
 		facial_lms shape = f_lms[idx];
-	//	for(uint8_t i = 0; i < 30; i++)
-	//	    cv::circle(frame, cv::Size(shape.part(i).x(), shape.part(i).y()), 
-	//	               2, CV_RGB(0, 255, 0), 2);
+		for(uint8_t i = 0; i < 26; i++)
+		    cv::circle(frame, cv::Size(shape.part(i).x(), shape.part(i).y()), 
+		               1.5, CV_RGB(0, 255, 0), 1.5);
 	}
-        //cv::imshow("face landmark", frame);
-	//if(27 == cv::waitKey(10))
-	//    break;
+        cv::imshow("face landmark", frame);
+	if(27 == cv::waitKey(10))
+	    break;
     	en_time = (float)((getTickCount() - st_time) / getTickFrequency());
 	printf("Frame #%d, Time: %f\n", n_f, en_time);
 	acc += en_time;

examples/sense/vision/FingersCountTest.cpp

@@ -31,7 +31,6 @@ int main(int argc, char **argv)
 
     cv::Mat frame, img, img2;
     std::vector<cv::Rect> hands;
-    std::vector<std::string> snum({"one", "two", "three", "four", "five", "si", "sev"});
 
     while (true)
     {
@@ -45,8 +44,8 @@ int main(int argc, char **argv)
 	                                hands[i].y - hands[i].y*SCALE,
 					hands[i].width + hands[i].width*SCALE,
 					hands[i].height + hands[i].height*SCALE));
-	    cv::putText(frame, snum[fc.num_fingers(img)], cv::Point(20, 20), 
-	              cv::FONT_HERSHEY_COMPLEX_SMALL, 0.8, CV_RGB(50, 50, 50), 2,
+	    cv::putText(frame, std::to_string(fc.num_fingers(img)), cv::Point(20, 20), 
+	              cv::FONT_HERSHEY_COMPLEX_SMALL, 1.5, CV_RGB(50, 50, 50), 2,
 		      CV_AA);
 	    cv::cvtColor(img, img2, COLOR_GRAY2BGR);
 	    for(size_t j = 0 ; j < fc.f_tips.size(); j++)

examples/sense/vision/VisionTestFace.cpp

@@ -12,6 +12,7 @@
 #include <iostream>
 #include <cstring>
 #include <vector>
+#include <signal.h>
 
 #include "sense/vision/CamCapture.hpp"
 #include "sense/vision/ITColor2Gray.hpp"

func_times

@@ -63,3 +63,19 @@ Face Landmark (26 parts)
   Average Time per Frame (no Face): 0.037780
   Average Time per Frame (on Face): 0.049697
   Average Time per Frame: 0.039970
+
+
+
+---------------------------------------------
+              [ All in One ]
+Face Landmark (26 parts) and FingersCount
+  Average Time per Frame (no Face): 0.031070
+  Average Time per Frame (on Face): 0.035642
+  Average Time per Frame (no Hand): 0.036242
+  Average Time per Frame (on Hand): 0.032650
+  Average Time per Frame: 0.034204
+
+
+
+Face Landmark (26 parts) 2 faces and FingersCount
+

include/sense/vision/FingersCount.hpp

@@ -30,8 +30,8 @@
   at current hand based on size or distance
   from other points or whatever
 */
-#define MIN_DEPTH 5.0
-#define MAX_DEPTH 50.0
+#define MIN_DEPTH 15.0
+#define MAX_DEPTH 70.0
 
 typedef std::vector<cv::Point> VP;
 typedef std::vector<VP> VVP;

sense/vision/detect/FingersCount.cpp

@@ -38,7 +38,7 @@ uint8_t FingersCount::num_fingers(cv::Mat hand)
 	cv::convexHull(cntr, hull, false);
 	if(hull.size() > 3)
 	    cv::convexityDefects(cntr, hull, defects);
-	count = 0;
+	count = 1;
 	VV4i::const_iterator d_it = defects.begin();
 	if(debug) printf("-----------------\n"); //XXX remove
 	while(d_it != defects.end())