Made same file for single images and multiple images in a .txt file

CMakeLists.txt

@@ -101,11 +101,8 @@ if(WITH_TOOLS)
 
     find_package(Boost COMPONENTS program_options REQUIRED)
 
-    add_executable(gazr_benchmark_head_pose_single_frame tools/benchmark_head_pose_estimation_single_frame.cpp)
-    target_link_libraries(gazr_benchmark_head_pose_single_frame gazr ${OpenCV_LIBRARIES})
-
-    add_executable(gazr_benchmark_head_pose_multiple_frames tools/benchmark_head_pose_estimation_multiple_frames.cpp)
-    target_link_libraries(gazr_benchmark_head_pose_multiple_frames gazr ${OpenCV_LIBRARIES})
+    add_executable(gazr_estimate_head_pose tools/estimate_head_pose_from_image_or_file.cpp)
+    target_link_libraries(gazr_estimate_head_pose gazr ${OpenCV_LIBRARIES})
 
     add_executable(gazr_estimate_head_direction tools/estimate_head_direction.cpp)
     target_link_libraries(gazr_estimate_head_direction gazr ${OpenCV_LIBRARIES} ${Boost_LIBRARIES})

README.md

@@ -69,15 +69,13 @@ $ make
 Run ``./gazr_show_head_pose ../share/shape_predictor_68_face_landmarks.dat`` to test
 the library. You should get something very similar to the picture above.
 
-#### Example - single frame
+#### Example - estimate head pose on image/images
 
-Run ``./gazr_benchmark_head_pose_single_frame ../share/shape_predictor_68_face_landmarks.dat frame.jpg``
+Run ``./gazr_estimate_head_pose ../share/shape_predictor_68_face_landmarks.dat frame.jpg``
 to print the head pose detected in _frame.jpg_.
 
-#### Example - multiple frames
-
-Run ``./gazr_benchmark_head_pose_multiple_frames ../share/shape_predictor_68_face_landmarks.dat filenames.txt``
-to print the head pose detected in each file listed in _filenames.txt_.
+Run ``./gazr_estimate_head_pose ../share/shape_predictor_68_face_landmarks.dat ex_images.txt``
+to print the head pose detected in each image file listed in _filenames.txt_ (image file names written in new lines).
 
 
 ### Installation - 2/2

tools/estimate_head_pose_from_image_or_file.cpp

@@ -0,0 +1,153 @@
+#define STR_EXPAND(tok) #tok
+#define STR(tok) STR_EXPAND(tok)
+
+#ifdef OPENCV3
+#include <opencv2/imgcodecs.hpp>
+#else
+#include <opencv2/highgui/highgui.hpp>
+#endif
+
+#include <iostream>
+
+#include "../src/head_pose_estimation.hpp"
+
+using namespace std;
+using namespace cv;
+
+const static size_t NB_TESTS = 100; // number of time the detection is run, to get better average detection duration
+
+std::vector<std::string> readFileToVector(const std::string& filename)
+{
+    std::ifstream source;
+    source.open(filename);
+    std::vector<std::string> lines;
+    std::string line;
+    while (std::getline(source, line))
+    {
+        lines.push_back(line);
+    }
+    return lines;
+}
+
+
+void estimate_head_pose_on_frameFileName(const std::string& frameFileName, HeadPoseEstimation estimator, std::vector<head_pose>& prev_poses, bool print_prev_poses)
+{
+    cout << "Estimating head pose on " << frameFileName << endl;
+#ifdef OPENCV3
+    Mat img = imread(frameFileName, IMREAD_COLOR);
+#else
+    Mat img = imread(frameFileName, CV_LOAD_IMAGE_COLOR);
+#endif
+
+    auto nbfaces = 0;
+
+    for(size_t i = 0; i < NB_TESTS; i++) {
+        estimator.update(img);
+    }
+
+    // auto t_detection = getTickCount();
+
+    for(size_t i = 0; i < NB_TESTS; i++) {
+        auto poses = estimator.poses();
+        nbfaces += poses.size(); // this is completly artifical: the only purpose is to make sure the compiler does not optimize away estimator.poses()
+    }
+
+    cout << "Found " << nbfaces/NB_TESTS << " face(s)" << endl;
+
+    auto poses = estimator.poses();
+    if (poses.size() > 0) {
+        for(auto pose : poses) {
+            cout << "Head pose: (" << pose(0,3) << ", " << pose(1,3) << ", " << pose(2,3) << ")" << endl;
+        }
+        prev_poses = poses;
+    }
+    else if (print_prev_poses) {
+        for(auto pose : prev_poses) {
+            cout << "Head pose: (" << pose(0,3) << ", " << pose(1,3) << ", " << pose(2,3) << ")" << endl;
+        }
+    }
+    else {
+        cout << "Head pose not calculated!" << endl;
+    }
+
+}
+
+
+int main(int argc, char **argv)
+{
+    Mat frame;
+
+    if(argc < 3) {
+        cerr << argv[0] << " " << STR(GAZR_VERSION) << "\n\nUsage: " 
+             << endl << argv[0] << " model.dat frame.{jpg|png}\n\nOR\n\n"
+             << endl << argv[0] << " model.dat filenames.txt" << endl;
+#ifdef HEAD_POSE_ESTIMATION_DEBUG
+        cerr <<  "Output: a new frame 'head_pose_<frame>.png'" << endl;
+#endif
+        return 1;
+    }
+
+    std::string fileName = argv[2];
+
+    if (fileName.find(".jpg") == std::string::npos and
+        fileName.find(".png") == std::string::npos and
+        fileName.find(".txt") == std::string::npos) {
+        cerr << "Please input a .jpg file, or .png file, or .txt file containing list of images in individual lines.";
+        return 1;
+    }
+
+    auto estimator = HeadPoseEstimation(argv[1]);
+    estimator.focalLength = 500;
+
+    cout << "Running " << NB_TESTS << " loops to get a good performance estimate..." << endl;
+#ifdef HEAD_POSE_ESTIMATION_DEBUG
+    cerr <<  "ATTENTION! The benchmark is compiled in DEBUG mode: the performance is no going to be good!!" << endl;
+#endif
+
+    // Prev pose (default)
+    head_pose prev_pose = {
+        -1,    -1,    -1,    -1.,
+        -1,    -1,    -1,    -1.,
+        -1,    -1,    -1,    -1.,
+        0,     0,     0,     1
+    };
+    std::vector<head_pose> prev_poses;
+    prev_poses.push_back(prev_pose);
+
+    auto t_start = getTickCount();
+
+    // Single image file
+    if (fileName.find("jpg") != std::string::npos or fileName.find("png") != std::string::npos) {
+        cout << ".jpg or .png file" << endl;
+        estimate_head_pose_on_frameFileName(fileName, estimator, prev_poses, false);
+    }
+
+    // Multiple lines
+    else {
+        cout << ".txt file" << endl;
+        // Read file
+        std::string frameFilesTxt(fileName);
+        std::vector<std::string> frameFileNames = readFileToVector(frameFilesTxt);
+
+        if (frameFileNames.size() == 0) {
+            cout << fileName << "does not exist, or has no image names" << endl;
+        }
+
+        for(auto frameFileName: frameFileNames) {
+            if (frameFileName.find("jpg") != std::string::npos or frameFileName.find("png") != std::string::npos) {
+                estimate_head_pose_on_frameFileName(frameFileName, estimator, prev_poses, false);
+            }
+        }
+    }
+
+    auto t_end = getTickCount();
+
+    // cout << "Face feature detection: " <<((t_detection-t_start) / NB_TESTS) /getTickFrequency() * 1000. << "ms;";
+    // cout << "Pose estimation: " <<((t_end-t_detection) / NB_TESTS) /getTickFrequency() * 1000. << "ms;";
+    cout << "Total time: " << (t_end-t_start) / getTickFrequency() * 1000. << "ms" << endl;
+
+#ifdef HEAD_POSE_ESTIMATION_DEBUG
+    imwrite("head_pose.png", estimator._debug);
+#endif
+
+}

tools/ex_images.txt

@@ -0,0 +1,3 @@
+/home/voletiv/AMOUNT_00745_19.jpg
+/home/voletiv/AMOUNT_00745_20.jpg
+/home/voletiv/AMOUNT_00745_21.jpg