opencv-126-DNN 基于残差网络的人脸检测

发表于 | 分类于

知识点

OpenCV在DNN模块中提供了基于残差SSD网络训练的人脸检测模型，该模型分别提供了tensorflow版本，caffe版本，torch版本模型文件，其中tensorflow版本的模型做了更加进一步的压缩优化，大小只有2MB左右，非常适合移植到移动端使用，实现人脸检测功能，而caffe版本的是fp16的浮点数模型，精准度更好。要先获得这些模型，只要下载OpenCV4.0源码之后，
打开运行sources\samples\dnn\face_detector\download_weights.py该脚本即可。
同样一张图像，在OpenCV HAAR与LBP级联检测器中必须通过不断调整参数才可以检测出全部人脸，而通过使用该模型，基本在Python语言中基于OpenCV后台的推断，在25毫秒均可以检测出结果，网络支持输入size大小为300x300。

代码（c++,python）

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include <iostream>

using namespace cv;
using namespace cv::dnn;
using namespace std;

const size_t width = 300;
const size_t height = 300;
String model_bin = "D:/projects/opencv_tutorial/data/models/face_detector/opencv_face_detector_uint8.pb";
String config_text = "D:/projects/opencv_tutorial/data/models/face_detector/opencv_face_detector.pbtxt";

int main(int argc, char** argv) {
	Mat frame = imread("D:/images/persons.png");
	if (frame.empty()) {
		printf("could not load image...\n");
		return -1;
	}
	namedWindow("input image", WINDOW_AUTOSIZE);
	imshow("input image", frame);

	Net net = readNetFromTensorflow(model_bin, config_text);
	net.setPreferableBackend(DNN_BACKEND_OPENCV);
	net.setPreferableTarget(DNN_TARGET_CPU);
	Mat blobImage = blobFromImage(frame, 1.0,
		Size(300, 300),
		Scalar(104.0, 177.0, 123.0), false, false);

	net.setInput(blobImage, "data");
	Mat detection = net.forward("detection_out");
	vector<double> layersTimings;
	double freq = getTickFrequency() / 1000;
	double time = net.getPerfProfile(layersTimings) / freq;
	printf("execute time : %.2f ms\n", time);

	Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr<float>());
	float confidence_threshold = 0.5;
	for (int i = 0; i < detectionMat.rows; i++) {
		float confidence = detectionMat.at<float>(i, 2);
		if (confidence > confidence_threshold) {
			size_t objIndex = (size_t)(detectionMat.at<float>(i, 1));
			float tl_x = detectionMat.at<float>(i, 3) * frame.cols;
			float tl_y = detectionMat.at<float>(i, 4) * frame.rows;
			float br_x = detectionMat.at<float>(i, 5) * frame.cols;
			float br_y = detectionMat.at<float>(i, 6) * frame.rows;

			Rect object_box((int)tl_x, (int)tl_y, (int)(br_x - tl_x), (int)(br_y - tl_y));
			rectangle(frame, object_box, Scalar(0, 0, 255), 2, 8, 0);
			putText(frame, format(" confidence %.2f", confidence),
				Point(tl_x - 10, tl_y - 5), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(255, 0, 0), 1, 8);
		}
	}
	imshow("ssd-face-detection", frame);

	waitKey(0);
	return 0;
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
"""
DNN 基于残差网络的人脸检测
"""

import cv2 as cv

model_bin ="opencv_face_detector_uint8.pb"
config_text = "opencv_face_detector.pbtxt"

# load tensorflow model
net = cv.dnn.readNetFromTensorflow(model_bin, config=config_text)
image = cv.imread("images/persons.jpg")
h = image.shape[0]
w = image.shape[1]

# 人脸检测
blobImage = cv.dnn.blobFromImage(image, 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())
cv.putText(image, label, (0, 15), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0))

# 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(image, (int(left), int(top)), (int(right), int(bottom)), (255, 0, 0), thickness=2)
        cv.putText(image, "score:%.2f"%score, (int(left), int(top)), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)

cv.imshow('face-detection-demo', image)

cv.waitKey(0)
cv.destroyAllWindows()

结果

代码地址

github