Mastering OpenCV 3 - Second Edition

Training and visualization

Unlike the other classes detailed in this chapter, training a face_tracker object does not involve any learning process. It is implemented in train_face_tracker.cpp simply as:

    face_tracker tracker; 
    tracker.smodel = load_ft<shape_model>(argv[1]); 
    tracker.pmodel = load_ft<patch_models>(argv[2]); 
    tracker.detector = load_ft<face_detector>(argv[3]); 
    save_ft<face_tracker>(argv[4],tracker);

Here arg[1] to argv[4] contain the paths to the shape_model, patch_model, face_detector, and face_tracker objects, respectively. The visualization for the face tracker in visualize_face_tracker.cpp is equally simple. Obtaining its input image stream either from a camera or video file, through the cv::VideoCapture class, the program ...