dlib系列笔记3：人脸检测

dlib系列笔记3：人脸检测

效果图

源码

截取重要类的实现片段
具体源码参考 https://github.com/lightjiang/FaceRecognize.git

• base.py

class Base(object):
    base_path = BASEPATH

    def __init__(self):
        self.img = None
        self.img_path = ''

    def load_img(self, path, relative=True, flags=cv2.IMREAD_COLOR):
        # 自动从web加载图片
        if path.startswith('http'):
            path = self.download_web_img(path)

        img_path = path
        if relative:
            path = os.path.join(self.base_path, path)
        if os.path.exists(path):
            res = cv2.imread(path, flags=flags)
            # 转换图片通道， opencv默认bgr格式，dlib需要rgb格式
            self.img = cv2.cvtColor(res, cv2.COLOR_BGR2RGB)
            self.img_path = img_path
            return self.img
        else:
            raise FileNotFoundError(path)

    def download_web_img(self, url):
        path = 'data/auto_download_img/%s.jpg' % uuid.uuid1()
        # 复制的百度图片链接有时会下载异常
        request.urlretrieve(url, path)
        print('download complete')
        return path

    def save_img(self, img, path, relative=True):
        if relative:
            path = os.path.join(self.base_path, path)
        cv2.imwrite(path, cv2.cvtColor(img, cv2.COLOR_RGB2BGR))

    def add_marks(self, pos, color=None):
        # 在self.img 上添加标记
        if isinstance(pos, tuple):
            pos = [pos]
        elif isinstance(pos, list):
            pos = pos
        else:
            raise AttributeError

        if not color:
            color = (0, 255, 0)
        for p in pos:
            cv2.circle(self.img, p, 2, color, 1)

    def add_faces(self, faces, show_score=True, color=None, add_text=None):
        # 在self.img 上添加脸部识别框
        if isinstance(faces, dict):
            faces = [faces]
        elif isinstance(faces, list):
            faces = faces
        else:
            raise AttributeError
        for face in faces:
            rect = face['position']
            if not color:
                color = (255, 0, 0)
            cv2.rectangle(self.img, (rect.left(), rect.top()), (rect.right(), rect.bottom()), color, 3)
            if show_score and 'score' in face:
                score = face['score']
                width = rect.right() - rect.left()
                cv2.putText(self.img, str(round(score, 3)), (rect.left() + 10, rect.bottom() - 10), cv2.FONT_HERSHEY_SIMPLEX, width/120,
                            (255, 255, 255), 1)
            if add_text:
                width = rect.right() - rect.left()
                cv2.putText(self.img, str(add_text), (rect.left() + 10, rect.bottom() - 10),
                            cv2.FONT_HERSHEY_SIMPLEX, width / 120,
                            (255, 255, 255), 1)

• face_detector.py

class FaceDetector(Base):
    """
    based on HOG
    """
    def __init__(self):
        self.face_detector = dlib.get_frontal_face_detector()
        super().__init__()

    def detect_face(self, log_status=True):
        if self.img is None:
            raise AttributeError('please load img before detect')
        t = time.time()
        dets, scores, idx = self.face_detector.run(self.img, 1, -1)
        res = []
        for index, face in enumerate(dets):
            temp = {
                'id': str(uuid.uuid1()),
                'position': face,
                'score': scores[index],
                'src': self.img_path
            }
            if scores[index] > 0:
                res.append(temp)
        if log_status:
            print("Detecting faces takes: {}s\nNumber of faces detected: {}".format(time.time() - t, len(res)))
        return res

    def detect_faces_from_imgs(self, imgs: list):
        t = time.time()
        res = []
        for img_path in imgs:
            self.load_img(img_path)
            res += self.detect_face(log_status=False)
        self.img = None
        print("Detecting faces takes: {}s\nNumber of faces detected: {}".format(time.time() - t, len(res)))
        return res

• cpp_src/face_detector.cpp

#include <dlib/image_processing/frontal_face_detector.h>
#include <dlib/gui_widgets.h>
#include <dlib/image_io.h>
#include <dlib/opencv.h>
#include <iostream>
#include <opencv2/core.hpp>
#include <opencv2/imgcodecs.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>
#include "mybase.hpp"

using namespace dlib;
using namespace std;

// ----------------------------------------------------------------------------------------


int main(int argc, char **argv)
{
    try
    {
        if (argc == 1)
        {
            cout << "Give some image files as arguments to this program." << endl;
            return 0;
        }

        frontal_face_detector detector = get_frontal_face_detector();

        // Loop over all the images provided on the command line.
        for (int i = 1; i < argc; ++i)
        {
            cout << "processing image " << argv[i] << endl;
            cv::Mat image;
            image = cv::imread(argv[i], cv::IMREAD_COLOR);
            dlib::cv_image<rgb_pixel> img(image);
            std::vector<rectangle> dets = detector(img);
            cout << "Number of faces detected: " << dets.size() << dets[0].left() <<dets[1] << image.size() << endl<< image;
            for (auto temp_point: dets){
                cout << temp_point<< temp_point.left() <<endl;
                mark_face(image, temp_point);
                // cv::rectangle(image, cv::Point(temp_point.left(), temp_point.top()), cv::Point(temp_point.right(), temp_point.bottom()),cv::Scalar(255,0,0),1,1,0);
            }
            // cv::rectangle(image, cv::Point(dets[0][0][0], dets[0][0][1]));
            cv::namedWindow("Display window", cv::WINDOW_AUTOSIZE); // Create a window for display.
            cv::imshow("Display window", image);                // Show our image inside it.
            cv::waitKey(0);                                     // Wait for a keystroke in the window
        }
    }
    catch (exception &e)
    {
        cout << "\nexception thrown!" << endl;
        cout << e.what() << endl;
    }
}