Python-opencv学习第三十课：人脸检测

Python-opencv学习第三十课：人脸检测

---

一、学习部分

记录笔者学习Python-opencv学习第三十课：人脸检测，代码资料来源于网络贾老师视频。

二、代码部分

1.引入库

代码如下：

import cv2 as cv
import matplotlib.pyplot as plt
import numpy as np

2.完整代码

代码如下：

import cv2 as cv
import numpy as np
from matplotlib import pyplot as plt


def read_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    # cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.imshow("input", image)
    cv.waitKey(0)
    cv.destroyAllWindows()


def color_space_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
    hsv = cv.cvtColor(image, cv.COLOR_HSV2BGR) # H 0 ~180
    # cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.imshow("gray", gray)
    cv.imshow("hsv", hsv)
    cv.waitKey(0)
    cv.destroyAllWindows()


def mat_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    h,w,c = image.shape
    roi = image[60:200, 60:280, :]
    blank = np.zeros((h, w, c), dtype=np.uint8)
    # blank[60:200, 60:280, :] = image[60:200, 60:280, :]
    blank = image # np.copy(image)
    cv.imshow("blank", blank)
    cv.imshow("roi", roi)
    cv.waitKey(0)
    cv.destroyAllWindows()


def pixel_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    cv.imshow("input", image)
    h,w,c = image.shape
    for row in range(h):
        for col in range(w):
            b,g,r = image[row, col]
            image[row, col] = (0, g, r)
    cv.imshow("result", image)
    cv.imwrite("D:/image_result.png", image)
    cv.waitKey(0)
    cv.destroyAllWindows()


def math_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    cv.imshow("input", image)
    h,w,c = image.shape
    blank = np.zeros_like(image)
    blank[:,:] = (2, 2, 2)
    cv.imshow("blank", blank)
    result = cv.multiply(image, blank)
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()

def nothing(x):
    print(x)

def adjust_lightness_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.createTrackbar("lightness", "input", 0, 100, nothing)
    cv.imshow("input", image)
    blank = np.zeros_like(image)
    while True:
        pos = cv.getTrackbarPos("lightness", "input")
        blank[:,:] = (pos, pos, pos)
        # cv.imshow("blank", blank)
        result = cv.add(image, blank)
        cv.imshow("result", result)
        c = cv.waitKey(1)
        if c == 27:
            break
    cv.destroyAllWindows()


def adjust_contrast_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.createTrackbar("lightness", "input", 0, 100, nothing)
    cv.createTrackbar("contrast", "input", 100, 200, nothing)
    cv.imshow("input", image)
    blank = np.zeros_like(image)
    while True:
        light = cv.getTrackbarPos("lightness", "input")
        contrast = cv.getTrackbarPos("contrast", "input") / 100
        print("light: ", light, "contrast: ", contrast)
        result = cv.addWeighted(image, contrast, blank, 0.5, light)
        cv.imshow("result", result)
        c = cv.waitKey(1)
        if c == 27:
            break
    cv.destroyAllWindows()


def keys_demo():
    image = cv.imread("D:/images/test.png") # BGR, 0~255
    cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.imshow("input", image)
    while True:
        c = cv.waitKey(1)
        if c == 49:  #1
            gray = cv.cvtColor(image, cv.COLOR_BGR2GRAY)
            cv.imshow("result", gray)
        if c == 50:  #2
            hsv = cv.cvtColor(image, cv.COLOR_BGR2HSV)
            cv.imshow("result", hsv)
        if c == 51:  # 3
            invert = cv.bitwise_not(image)
            cv.imshow("result", invert)
        if c == 27:
            break
    cv.destroyAllWindows()


def color_table_demo():
    colormap = [
        cv.COLORMAP_AUTUMN,
        cv.COLORMAP_BONE,
        cv.COLORMAP_JET,
        cv.COLORMAP_WINTER,
        cv.COLORMAP_RAINBOW,
        cv.COLORMAP_OCEAN,
        cv.COLORMAP_SUMMER,
        cv.COLORMAP_SPRING,
        cv.COLORMAP_COOL,
        cv.COLORMAP_PINK,
        cv.COLORMAP_HOT,
        cv.COLORMAP_PARULA,
        cv.COLORMAP_MAGMA,
        cv.COLORMAP_INFERNO,
        cv.COLORMAP_PLASMA,
        cv.COLORMAP_VIRIDIS,
        cv.COLORMAP_CIVIDIS,
        cv.COLORMAP_TWILIGHT,
        cv.COLORMAP_TWILIGHT_SHIFTED ]

    image = cv.imread("D:/images/canjian.jpg") # BGR, 0~255
    cv.namedWindow("input", cv.WINDOW_AUTOSIZE)
    cv.imshow("input", image)
    index = 0
    while True:
        dst = cv.applyColorMap(image, colormap[index%19])
        index += 1
        cv.imshow("color style", dst);
        c = cv.waitKey(400)
        if c == 27:
            break
    cv.destroyAllWindows()


def bitwise_demo():
    b1 = np.zeros((400, 400, 3), dtype=np.uint8)
    b1[:,:] = (255, 0, 255)
    b2 = np.zeros((400, 400, 3), dtype=np.uint8)
    b2[:,:] = (0, 255, 255)
    cv.imshow("b1", b1);
    cv.imshow("b2", b2);

    dst1 = cv.bitwise_and(b1, b2)
    dst2 = cv.bitwise_or(b1, b2)
    cv.imshow("bitwise_and", dst1)
    cv.imshow("bitwise_or", dst2)
    cv.waitKey(0)
    cv.destroyAllWindows()


def channels_split_demo():
    b1 = cv.imread("D:/images/lena.jpg")
    print(b1.shape)
    cv.imshow("input", b1)
    cv.imshow("b1", b1[:,:,2])
    mv = cv.split(b1)
    mv[0][:,:] = 255
    result = cv.merge(mv)

    dst = np.zeros(b1.shape, dtype=np.uint8)
    cv.mixChannels([b1], [dst], fromTo=[2, 0, 1, 1, 0, 2])
    cv.imshow("output4", dst)

    cv.imshow("result",result)
    cv.waitKey(0)
    cv.destroyAllWindows()


def color_space_demo():
    b1 = cv.imread("D:/images/greenback.png")
    print(b1.shape)
    cv.imshow("input", b1)
    hsv = cv.cvtColor(b1, cv.COLOR_BGR2HSV)
    cv.imshow("hsv", hsv)
    mask = cv.inRange(hsv, (35, 43, 46), (77, 255, 255))
    cv.bitwise_not(mask, mask)
    result = cv.bitwise_and(b1, b1, mask=mask)
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()


def pixel_stat_demo():
    b1 = cv.imread("D:/images/1024.png")
    print(b1.shape)
    cv.imshow("input", b1)
    print(np.max(b1[:,:,2]))
    means, dev = cv.meanStdDev(b1)
    print(means, "dev: ", dev)
    cv.waitKey(0)
    cv.destroyAllWindows()


def drawing_demo():
    b1 = cv.imread("D:/images/1024.png")#np.zeros((512, 512, 3), dtype=np.uint8)
    temp = np.copy(b1)
    cv.rectangle(b1, (50, 50), (400, 400), (0, 0, 255), 4, 8, 0)
    # cv.circle(b1, (200, 200), 100, (255, 0, 0), -1, 8, 0)
    # cv.line(b1, (50, 50), (400, 400), (0, 255, 0), 4, 8, 0)
    cv.putText(b1, "99% face", (50, 50), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 255, 255), 2, 8)
    cv.imshow("input", b1)
    cv.waitKey(0)
    cv.destroyAllWindows()


def random_color_demo():
    b1 = np.zeros((512, 512, 3), dtype=np.uint8)
    while True:
        xx = np.random.randint(0, 512, 2, dtype=np.int)
        yy = np.random.randint(0, 512, 2, dtype=np.int)
        bgr = np.random.randint(0, 255, 3, dtype=np.int32)
        print(bgr[0], bgr[1], bgr[2])
        cv.line(b1, (xx[0], yy[0]), (xx[1], yy[1]), (np.int(bgr[0]), np.int(bgr[1]), np.int(bgr[2])), 1, 8, 0)
        cv.imshow("input", b1)
        c = cv.waitKey(10)
        if c == 27:
            break
    cv.destroyAllWindows()


def polyline_drawing_demo():
    canvas = np.zeros((512, 512, 3), dtype=np.uint8)
    pts = np.array([[100, 100], [350, 100], [450, 280], [320, 450], [80, 400]],  dtype=np.int32)
    # cv.fillPoly(canvas, [pts], (255, 0, 255), 8, 0);
    # cv.polylines(canvas, [pts], True, (0, 0, 255), 2, 8, 0);
    cv.drawContours(canvas, [pts], -1, (255, 0, 0), -1);
    cv.imshow("polyline", canvas);
    cv.waitKey(0)
    cv.destroyAllWindows()


b1 = cv.imread("D:/images/1024.png")  # np.zeros((512, 512, 3), dtype=np.uint8)
img = np.copy(b1)
x1 = -1
x2 = -1
y1 = -1
y2 = -1
def mouse_drawing(event, x, y, flags, param):
    global x1, y1, x2, y2
    if event == cv.EVENT_LBUTTONDOWN:
        x1 = x
        y1 = y
    if event == cv.EVENT_MOUSEMOVE:
        if x1 < 0 or y1 < 0:
            return
        x2 = x
        y2 = y
        dx = x2 - x1
        dy = y2 - y1
        if dx > 0 and dy > 0:
            b1[:,:,:] = img[:,:,:]
            cv.rectangle(b1, (x1, y1), (x2, y2), (0, 0, 255), 2, 8, 0)
    if event == cv.EVENT_LBUTTONUP:
        x2 = x
        y2 = y
        dx = x2 - x1
        dy = y2 - y1
        if dx > 0 and dy > 0:
            b1[:, :, :] = img[:,:,:]
            cv.rectangle(b1, (x1, y1), (x2, y2), (0, 0, 255), 2, 8, 0)
        x1 = -1
        x2 = -1
        y1 = -1
        y2 = -1

def mouse_demo():
    cv.namedWindow("mouse_demo", cv.WINDOW_AUTOSIZE)
    cv.setMouseCallback("mouse_demo", mouse_drawing)
    while True:
        cv.imshow("mouse_demo", b1)
        c = cv.waitKey(10)
        if c == 27:
            break
    cv.destroyAllWindows()


def norm_demo():
    image = cv.imread("D:/images/1024.png")
    cv.namedWindow("norm_demo", cv.WINDOW_AUTOSIZE)
    result = np.zeros_like(np.float32(image))
    cv.normalize(np.float32(image), result, 0, 1, cv.NORM_MINMAX, dtype=cv.CV_32F)
    cv.imshow("norm_demo", np.uint8(result*255))
    cv.waitKey(0)
    cv.destroyAllWindows()


def resize_demo():
    image = cv.imread("D:/images/1024.png")
    h, w, c = image.shape
    cv.namedWindow("resize", cv.WINDOW_AUTOSIZE)
    dst = cv.resize(image, (0, 0), fx=0.75, fy=0.75, interpolation=cv.INTER_NEAREST)
    cv.imshow("resize", dst)
    cv.waitKey(0)
    cv.destroyAllWindows()


def flip_demo():
    image = cv.imread("D:/images/1024.png")
    cv.imshow("input", image)
    cv.namedWindow("flip", cv.WINDOW_AUTOSIZE)
    dst = cv.flip(image, -1)
    cv.imshow("flip", dst)
    cv.waitKey(0)
    cv.destroyAllWindows()


def rotate_demo():
    src = cv.imread("D:/images/dannis1.png")
    cv.imshow("input", src)
    h, w, c = src.shape
    M = np.zeros((2, 3), dtype=np.float32)
    alpha = np.cos(np.pi / 4.0)
    beta = np.sin(np.pi / 4.0)
    print("alpha : ", alpha)

    # 初始旋转矩阵
    M[0, 0] = alpha
    M[1, 1] = alpha
    M[0, 1] = beta
    M[1, 0] = -beta
    cx = w / 2
    cy = h / 2
    tx = (1 - alpha) * cx - beta * cy
    ty = beta * cx + (1 - alpha) * cy
    M[0, 2] = tx
    M[1, 2] = ty

    # change with full size
    bound_w = int(h * np.abs(beta) + w * np.abs(alpha))
    bound_h = int(h * np.abs(alpha) + w * np.abs(beta))

    # 添加中心位置迁移
    M[0, 2] += bound_w / 2 - cx
    M[1, 2] += bound_h / 2 - cy
    dst = cv.warpAffine(src, M, (bound_w, bound_h))
    cv.imshow("rotate without cropping", dst)
    cv.waitKey(0)
    cv.destroyAllWindows()


def video_demo():
    cap = cv.VideoCapture("D:/images/video/dushuhu.mp4")
    w = cap.get(cv.CAP_PROP_FRAME_WIDTH)
    h = cap.get(cv.CAP_PROP_FRAME_HEIGHT)
    fps = cap.get(cv.CAP_PROP_FPS)
    out = cv.VideoWriter("D:/test.mp4",  cv.CAP_ANY, np.int(cap.get(cv.CAP_PROP_FOURCC)), fps, (np.int(w), np.int(h)), True)
    print(w, h, fps)
    while True:
        ret, frame = cap.read()
        if ret is not True:
            break
        cv.imshow("frame", frame)
        hsv = cv.cvtColor(frame, cv.COLOR_BGR2HSV)
        cv.imshow("result", hsv)
        out.write(hsv)
        c = cv.waitKey(10)
        if c == 27:
            break
    cv.destroyAllWindows()

    out.release()
    cap.release()


def image_hist():
    image = cv.imread("D:/images/dannis1.png")
    cv.imshow("input", image)
    color = ('blue', 'green', 'red')
    for i, color in enumerate(color):
        hist = cv.calcHist([image], [i], None, [32], [0, 256])
        print(hist)
        plt.plot(hist, color=color)
        plt.xlim([0, 32])
    plt.show()
    cv.waitKey(0)
    cv.destroyAllWindows()


def hist2d_demo():
    image = cv.imread("D:/images/dannis1.png")
    hsv = cv.cvtColor(image, cv.COLOR_BGR2HSV)
    hist = cv.calcHist([hsv], [0, 1], None, [48, 48], [0, 180, 0, 256])
    dst = cv.resize(hist, (400, 400))
    cv.normalize(dst, dst, 0, 255, cv.NORM_MINMAX)
    cv.imshow("image", image)
    dst = cv.applyColorMap(np.uint8(dst), cv.COLORMAP_JET)
    cv.imshow("hist", dst)
    plt.imshow(hist, interpolation='nearest')
    plt.title("2D Histogram")
    plt.show()
    cv.waitKey(0)
    cv.destroyAllWindows()


def eqhist_demo():
    image = cv.imread("D:/images/cat.jpg", cv.IMREAD_GRAYSCALE)
    cv.imshow("input", image)
    result = cv.equalizeHist(image)
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()

def blur_demo():
    image = cv.imread("D:/images/cat.jpg")
    cv.imshow("input", image)
    result = cv.blur(image, (15, 15))
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()


def conv_demo():
    image = cv.imread("D:/images/cat.jpg")
    cv.imshow("input", image)
    result = cv.GaussianBlur(image, (0, 0), 15)
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()


def bifilter_demo():
    image = cv.imread("D:/images/example.png")
    cv.imshow("input", image)
    result = cv.bilateralFilter(image, 0, 100, 10)
    cv.imshow("result", result)
    cv.waitKey(0)
    cv.destroyAllWindows()


model_bin = "E:/python/opencv/opencv_face_detector_uint8.pb"
config_text = "E:/python/opencv/opencv_face_detector.pbtxt";
def face_detection_demo():
    net = cv.dnn.readNetFromTensorflow(model=model_bin, config=config_text)
    cap = cv.VideoCapture("C:/Users/akaak/Videos/01.mp4")
    while True:
        ret, frame = cap.read()
        h, w, c = frame.shape
        if ret is not True:
            break
        # NCHW
        blob = cv.dnn.blobFromImage(frame, 1.0, (300, 300), (104.0, 177.0, 123.0), False, False)
        net.setInput(blob)
        outs = net.forward() # 1x1xNx7
        for detection in outs[0, 0, :, :]:
            score = float(detection[2])
            if score > 0.5:
                left = detection[3] * w
                top = detection[4] * h
                right = detection[5] * w
                bottom = detection[6] * h
                cv.rectangle(frame, (np.int(left), np.int(top)), (np.int(right), np.int(bottom)), (0, 0, 255), 2, 8, 0)
        cv.imshow("frame", frame)
        c = cv.waitKey(1)
        if c == 27:
            break
    cv.destroyAllWindows()
    cap.release()


if __name__ == "__main__":
    face_detection_demo()

face_detector代码如下：

import cv2 as cv

model_bin = "E:/python/opencv/opencv_face_detector_uint8.pb"
config_text = "E:/python/opencv/opencv_face_detector.pbtxt";


def video_detection():
    # load tensorflow model
    net = cv.dnn.readNetFromTensorflow(model_bin, config=config_text)
    capture = cv.VideoCapture("C:/Users/akaak/Videos/01.mp4")

    # 人脸检测
    while True:
        e1 = cv.getTickCount()
        ret, frame = capture.read()
        if ret is not True:
            break
        h, w, c = frame.shape
        blobImage = cv.dnn.blobFromImage(frame, 1.0, (300, 300), (104.0, 177.0, 123.0), False, False);
        net.setInput(blobImage)
        cvOut = net.forward()

        # Put efficiency information.
        t, _ = net.getPerfProfile()
        label = 'Inference time: %.2f ms' % (t * 1000.0 / cv.getTickFrequency())

        # 绘制检测矩形
        for detection in cvOut[0,0,:,:]:
            score = float(detection[2])
            objIndex = int(detection[1])
            if score > 0.5:
                left = detection[3]*w
                top = detection[4]*h
                right = detection[5]*w
                bottom = detection[6]*h

                # 绘制
                cv.rectangle(frame, (int(left), int(top)), (int(right), int(bottom)), (255, 0, 0), thickness=2)
                cv.putText(frame, "score:%.2f"%score, (int(left), int(top)), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)
        e2 = cv.getTickCount()
        fps = cv.getTickFrequency() / (e2 - e1)
        cv.putText(frame, label + (" FPS: %.2f"%fps), (10, 50), cv.FONT_HERSHEY_SIMPLEX, 1.0, (0, 0, 255), 2)
        cv.imshow('face-detection-demo', frame)
        c = cv.waitKey(1)
        if c == 27:
            break
    cv.destroyAllWindows()


if __name__ == "__main__":
    video_detection()

三、运行结果

总结

本文介绍了笔者学习Python-opencv第三十课：人脸识别，感谢bilibili贾老师的视频，完成了大作业一部分任务，giao！（代码资料来源于网络贾老师视频）