【dlib代码解读】人脸检测器的训练【转】

版权声明：本文为博主原创文章，能帮到你很开心（^_^）~，欢迎转载，转载请注明出处~ https://blog.csdn.net/qq_36396104/article/details/79860189

上一篇：图像处理入门（一）：linux（ubuntu）配置Openface+测试

一、需要了解的基础知识：

    （1）使用Opencv对图片进行读写和显示操作：

               1、使用cv2.cv的LoadImage、ShowImage和SaveImage函数：

        import cv2.cv as cv

        # 读图片
        image=cv.LoadImage('img/image.png', cv.CV_LOAD_IMAGE_COLOR)#Load the image
        #Or just: image=cv.LoadImage('img/image.png')

        cv.NamedWindow('a_window', cv.CV_WINDOW_AUTOSIZE) #Facultative
        cv.ShowImage('a_window', image) #Show the image

        # 写图片
        cv.SaveImage("thumb.png", thumb)
        cv.WaitKey(0) #Wait for user input and quit

              2、也可以直接使用cv2的imread、imwrite和imshow函数

        import numpy as np
        import cv2

        img = cv2.imread('messi5.jpg',0)
        cv2.imshow('image',img)
        k = cv2.waitKey(0)
        if k == 27:         # wait for ESC key to exit
            cv2.destroyAllWindows()
        elif k == ord('s'): # wait for 's' key to save and exit
            cv2.imwrite('messigray.png',img)
            cv2.destroyAllWindows()
        imread函数还可以定义加载的mode，默认是以RGB模式处理图片：

        import cv2
        grayImage = cv2.imread('MyPic.png', cv2.CV_LOAD_IMAGE_GRAYSCALE)
        # 可选参数CV_LOAD_IMAGE_COLOR (BGR), CV_LOAD_IMAGE_GRAYSCALE (grayscale), CV_LOAD_IMAGE_UNCHANGED(neither)
        cv2.imwrite('MyPicGray.png', grayImage)

    （2）图像仿射：（引用：https://blog.csdn.net/on2way/article/details/46801063）

图像的旋转加上拉升就是图像仿射变换，仿射变化也是需要一个M矩阵就可以，但是由于仿射变换比较复杂，一般直接找很难找到这个矩阵，opencv提供了根据变换前后三个点的对应关系来自动求解M。这个函数是 
M=cv2.getAffineTransform(pos1,pos2),其中两个位置就是变换前后的对应位置关系。输 出的就是仿射矩阵M。然后在使用函数cv2.warpAffine()。形象化的图如下（引用参考的） 

                     

一个例子比如：

    import cv2
    import numpy as np
    import matplotlib.pyplot as plt
    img = cv2.imread('flower.jpg')
    pts1 = np.float32([[50,50],[200,50],[50,200]])
    rows,cols = img.shape[:2]
    M = cv2.getAffineTransform(pts1,pts2)
    pts2 = np.float32([[10,100],[200,50],[100,250]])
    #第三个参数：变换后的图像大小
    plt.subplot(122)
    res = cv2.warpAffine(img,M,(rows,cols))
    plt.subplot(121)    plt.imshow(img)
    plt.imshow(res)

二、人脸检测和处理：

        （一）、识别最大的人脸边界框：

        （1）、使用人脸检测器：dlib自带的frontal_face_detector特征提取器：

                    使用方法：

             detector=dlib.get_frontal_face_detector()
             faces = detector(Img, 1) 
             或者：
             dets = dlib.get_frontal_face_detector(img)

              具体的原理：

【dlib代码解读】人脸检测器的训练【转】

        （二）、人脸对齐：

         （1）、利用dlib的特征预测器（和dlib的shape_predictor_68_face_landmarks.dat（68点特征预测器）），进行人脸  68点特征提取：

        dlib人脸关键点检测器训练

        使用方法：

    predictor = dlib.shape_predictor("../models/dlib/shape_predictor_68_face_landmarks.dat")#此处填的是路径
    shape = predictor(img, dets[0])

三、代码和效果：

import dlib

import cv2


def GetFace(imgpath='1.jpg'):
    #读取图片
    Img=cv2.imread(imgpath)
    if Img is None:
        raise Exception("Unable to load image: {}".format(imgpath))
    # cv2.imshow('face image',Img);
    # cv2.waitKey()
    detector = dlib.get_frontal_face_detector()
    facebox = detector(Img, 1)
    face=facebox[0]#注意facebox是一个array
    predictor = dlib.shape_predictor("/home/angelo/openface/demos/../models/dlib/shape_predictor_68_face_landmarks.dat")  # 此处填的是路径
    shape = predictor(Img, face)
    # 生成dlib的图像窗口
    win = dlib.image_window()
    win.clear_overlay()
    win.set_image(Img)
    # 绘制面部轮廓
    win.add_overlay(shape)
    # 绘制矩阵轮廓
    win.add_overlay(facebox)
    # 保持图像
    dlib.hit_enter_to_continue()

if __name__=="__main__":
    GetFace('1.jpg')

图像的仿射：

def GetFace(imgpath='1.jpg'):
    #读取图片
    Img=cv2.imread(imgpath)
    if Img is None:
        raise Exception("Unable to load image: {}".format(imgpath))
    # cv2.imshow('face image',Img);
    # cv2.waitKey()
    detector = dlib.get_frontal_face_detector()
    facebox = detector(Img, 1)
    face=facebox[0]#注意facebox是一个array
    predictor = dlib.shape_predictor("/home/angelo/openface/demos/../models/dlib/shape_predictor_68_face_landmarks.dat")  # 此处填的是路径
    shape = predictor(Img, face)
    # 生成dlib的图像窗口
    win = dlib.image_window()
    win.clear_overlay()
    win.set_image(Img)
    # 绘制面部轮廓
    win.add_overlay(shape)
    # 绘制矩阵轮廓
    win.add_overlay(facebox)
    # 保持图像
    # dlib.hit_enter_to_continue()

    #图像仿射
    landmarkIndices = [39, 42, 57]
    npLandmarks = np.float32(list(map(lambda p: (p.x, p.y), shape.parts())))
    npLandmarkIndices = np.array(landmarkIndices)

    TEMPLATE = np.float32([
        (0.0792396913815, 0.339223741112), (0.0829219487236, 0.456955367943),
        (0.0967927109165, 0.575648016728), (0.122141515615, 0.691921601066),
        (0.168687863544, 0.800341263616), (0.239789390707, 0.895732504778),
        (0.325662452515, 0.977068762493), (0.422318282013, 1.04329000149),
        (0.531777802068, 1.06080371126), (0.641296298053, 1.03981924107),
        (0.738105872266, 0.972268833998), (0.824444363295, 0.889624082279),
        (0.894792677532, 0.792494155836), (0.939395486253, 0.681546643421),
        (0.96111933829, 0.562238253072), (0.970579841181, 0.441758925744),
        (0.971193274221, 0.322118743967), (0.163846223133, 0.249151738053),
        (0.21780354657, 0.204255863861), (0.291299351124, 0.192367318323),
        (0.367460241458, 0.203582210627), (0.4392945113, 0.233135599851),
        (0.586445962425, 0.228141644834), (0.660152671635, 0.195923841854),
        (0.737466449096, 0.182360984545), (0.813236546239, 0.192828009114),
        (0.8707571886, 0.235293377042), (0.51534533827, 0.31863546193),
        (0.516221448289, 0.396200446263), (0.517118861835, 0.473797687758),
        (0.51816430343, 0.553157797772), (0.433701156035, 0.604054457668),
        (0.475501237769, 0.62076344024), (0.520712933176, 0.634268222208),
        (0.565874114041, 0.618796581487), (0.607054002672, 0.60157671656),
        (0.252418718401, 0.331052263829), (0.298663015648, 0.302646354002),
        (0.355749724218, 0.303020650651), (0.403718978315, 0.33867711083),
        (0.352507175597, 0.349987615384), (0.296791759886, 0.350478978225),
        (0.631326076346, 0.334136672344), (0.679073381078, 0.29645404267),
        (0.73597236153, 0.294721285802), (0.782865376271, 0.321305281656),
        (0.740312274764, 0.341849376713), (0.68499850091, 0.343734332172),
        (0.353167761422, 0.746189164237), (0.414587777921, 0.719053835073),
        (0.477677654595, 0.706835892494), (0.522732900812, 0.717092275768),
        (0.569832064287, 0.705414478982), (0.635195811927, 0.71565572516),
        (0.69951672331, 0.739419187253), (0.639447159575, 0.805236879972),
        (0.576410514055, 0.835436670169), (0.525398405766, 0.841706377792),
        (0.47641545769, 0.837505914975), (0.41379548902, 0.810045601727),
        (0.380084785646, 0.749979603086), (0.477955996282, 0.74513234612),
        (0.523389793327, 0.748924302636), (0.571057789237, 0.74332894691),
        (0.672409137852, 0.744177032192), (0.572539621444, 0.776609286626),
        (0.5240106503, 0.783370783245), (0.477561227414, 0.778476346951)])

    TPL_MIN, TPL_MAX = np.min(TEMPLATE, axis=0), np.max(TEMPLATE, axis=0)
    MINMAX_TEMPLATE = (TEMPLATE - TPL_MIN) / (TPL_MAX - TPL_MIN)
    H = cv2.getAffineTransform(npLandmarks[npLandmarkIndices],
                               96 * MINMAX_TEMPLATE[npLandmarkIndices])  # 其中两个位置就是变换前后的对应位置关系。输 出的就是仿射矩阵M
    thumbnail = cv2.warpAffine(Img, H, (96, 96))  # 仿射函数cv2.warpAffine()接受三个参数，需要变换的原始图像，移动矩阵M
    # 以及变换的图像大小（这个大小如果不和原始图像大小相同，那么函数会自 动通过插值来调整像素间的关系）

    cv2.imshow('image', thumbnail);
    cv2.waitKey();

效果：96x96:

下一篇：图像处理（三）：在Windows系统里配置dlib环境并做图像批量处理