1基于CNN的性别分类建模原理

1.1 人脸识别

1.2 性别预测

1.3 年龄预测

1.4 结果

2 代码

参考

本教程中，我们将讨论应用于面部的深层学习的有趣应用。我们将估计年龄，并从单个图像中找出该人的性别。模型由GilLevi和TalHassner训练（https://talhassner.github.io/home/publication/2015_CVPR）。本文介绍了如何在OpenCV中使用该模型的步骤说明。Opencv版本3.4.3以上。代码教程代码可以分为四个部分：

1基于CNN的性别分类建模原理

作者使用非常简单的卷积神经网络结构，类似于Caffenet和Alexnet。网络使用3个卷积层、2个全连接层和一个最终的输出层。下面给出了这些层的细节。COV1：第一卷积层具有96个内核大小7的节点。COV2：第二个卷积层Conv层具有256个具有内核大小5的节点。CONV3：第三个CONV层具有384个内核大小为3的节点。两个完全连接的层各自具有512个节点。

训练数据来源：https://talhassner.github.io/home/projects/Adience/Adience-data.html

检测程序主要有四块：检测人脸检测、性别检测、年龄显示和输出。

1.1 人脸识别

我们将使用人脸检测器（tensorflow模型）进行人脸检测。该模型很简单，即使在CPU上也是相当快的。详细见论文：

https://arxiv.org/pdf/1502.00046.pdf

1.2 性别预测

将性别预测设定为一个分类问题。性别预测网络（caffe模型）中的输出层类型为两类，2个节点表示“男性”和“女性”两类。以这两个输出的最大值作为最终的性别。

1.3 年龄预测

理想情况下，年龄预测应该作为一个回归问题来处理。然而通过回归准确估计年龄是很有挑战性的。即使是人类也无法通过观察一个人来准确预测年龄。但是我们能够知道他们是20多岁还是30多岁。由于这个原因，把这个问题描述为一个分类问题是明智的，因为我们试图估计这个人所处的年龄组。例如，0-2范围内的年龄是一个类，4-6是另一个类，依此类推。因此数据集分为以下8个年龄组[(0-2)、(4-6)、(8-12)、(15-20)、(25-32)、(38-43)、(48-53)、(60-100)]。因此，年龄预测网络在最后一层有8个节点，表示所述年龄范围。

应该记住，从一幅图像中预测年龄并不是一个很容易解决的问题，因为感知到的年龄取决于许多因素，而同龄的人在世界各地可能看起来很不一样。而且，人们非常努力地隐藏他们的真实年龄！

我们加载年龄网络（caffe模型）并使用前向通道获得输出。由于网络结构类似于性别网络，所以我们可以从所有输出中提取出最大值来得到预测的年龄组

1.4 结果

尽管性别预测网络表现良好，但年龄预测网络仍未达到我们的预期。所以添加人脸对齐算法或者数据样本很多时候，可以通过回归的模型来检测。但是性别人脸检测还是很准确的。

2 代码

在VS2017下运行了C++代码，其中OpenCV版本至少要3.4.5以上。不然模型读取会有问题。三个模型文件太大，见下载链接：

https://download.csdn.net/download/luohenyj/10993309

https://github.com/luohenyueji/OpenCV-Practical-Exercise

如果没有积分（系统自动设定资源分数）看看参考链接。我搬运过来的，大修改没有。

其中tensorflow和caffe模型都可以用opencv中的readnet函数读取，流程很简单。看看代码就会。

代码提供了C++和Python版本，但是python版本没有运行，原因opencv版本太低，不想升级。代码都有详细的注释。

C++版本：


#include <tuple>#include <iostream>#include <opencv2/opencv.hpp>#include <opencv2/dnn.hpp>#include <iterator>using namespace cv;using namespace cv::dnn;using namespace std;/*** @brief Get the Face Box object 人脸定位** @param net 人脸检测网络* @param frame 检测图像* @param conf_threshold 阈值* @return tuple<Mat, vector<vector<int>>> 元组容器，可返回多个值*/tuple<Mat, vector<vector<int>>> getFaceBox(Net net, Mat &frame, double conf_threshold){//图像复制Mat frameOpenCVDNN = frame.clone();int frameHeight = frameOpenCVDNN.rows;int frameWidth = frameOpenCVDNN.cols;//缩放尺寸double inScaleFactor = 1.0;//检测图大小Size size = Size(300, 300);// std::vector<int> meanVal = {104, 117, 123};Scalar meanVal = Scalar(104, 117, 123);cv::Mat inputBlob;inputBlob = cv::dnn::blobFromImage(frameOpenCVDNN, inScaleFactor, size, meanVal, true, false);net.setInput(inputBlob, "data");//四维矩阵输出cv::Mat detection = net.forward("detection_out");//提取结果信息cv::Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr<float>());vector<vector<int>> bboxes;for (int i = 0; i < detectionMat.rows; i++){//预测概率float confidence = detectionMat.at<float>(i, 2);if (confidence > conf_threshold){//左上角点，坐标被归一化int x1 = static_cast<int>(detectionMat.at<float>(i, 3) * frameWidth);int y1 = static_cast<int>(detectionMat.at<float>(i, 4) * frameHeight);//右下角角点，坐标被归一化int x2 = static_cast<int>(detectionMat.at<float>(i, 5) * frameWidth);int y2 = static_cast<int>(detectionMat.at<float>(i, 6) * frameHeight);vector<int> box = { x1, y1, x2, y2 };//人脸坐标bboxes.push_back(box);//图像框选cv::rectangle(frameOpenCVDNN, cv::Point(x1, y1), cv::Point(x2, y2), cv::Scalar(0, 255, 0), 2, 4);}}return make_tuple(frameOpenCVDNN, bboxes);}int main(void){//人脸模型string faceProto = "model/opencv_face_detector.pbtxt";string faceModel = "model/opencv_face_detector_uint8.pb";//年龄模型string ageProto = "model/age_deploy.prototxt";string ageModel = "model/age_net.caffemodel";//性别模型string genderProto = "model/gender_deploy.prototxt";string genderModel = "model/gender_net.caffemodel";//均值Scalar MODEL_MEAN_VALUES = Scalar(78.4263377603, 87.7689143744, 114.895847746);//年龄段标签vector<string> ageList = { "(0-2)", "(4-6)", "(8-12)", "(15-20)", "(25-32)","(38-43)", "(48-53)", "(60-100)" };//性别标签vector<string> genderList = { "Male", "Female" };//导入网络Net ageNet = cv::dnn::readNet(ageProto, ageModel);Net genderNet = cv::dnn::readNet(genderProto, genderModel);Net faceNet = cv::dnn::readNetFromTensorflow(faceModel, faceProto);//打开摄像头VideoCapture cap;cap.open(0);if (cap.isOpened()){cout << "camera is opened!" << endl;}else{return 0;}int padding = 20;while (waitKey(1) < 0){// read frame 读图Mat frame;cap.read(frame);if (frame.empty()){waitKey();break;}frame = imread("./images/couple1.jpg");//人脸坐标vector<vector<int>> bboxes;//人脸检测结果图Mat frameFace;//人脸定位//tie()函数解包frameFace和bboxestie(frameFace, bboxes) = getFaceBox(faceNet, frame, 0.7);//人脸判断if (bboxes.size() == 0){cout << "No face detected, checking next frame." << endl;continue;}//逐个提取人脸检测for (auto it = begin(bboxes); it != end(bboxes); ++it){//框选人脸Rect rec(it->at(0) - padding, it->at(1) - padding, it->at(2) - it->at(0) + 2 * padding, it->at(3) - it->at(1) + 2 * padding);//避免人脸框选超过图像边缘rec.width = ((rec.x + rec.width) > frame.cols) ? (frame.cols - rec.x - 1) : rec.width;rec.height = ((rec.y + rec.height) > frame.rows) ? (frame.rows - rec.y - 1) : rec.height;// take the ROI of box on the frame,原图中提取人脸Mat face = frame(rec);//性别检测Mat blob;blob = blobFromImage(face, 1, Size(227, 227), MODEL_MEAN_VALUES, false);genderNet.setInput(blob);// string gender_preds; 获取前向传播softmax结果vector<float> genderPreds = genderNet.forward();// find max element index max_element用于找寻最大值// distance function does the argmax() work in C++ distance返回最大值和第一个值下标的距离int max_index_gender = std::distance(genderPreds.begin(), max_element(genderPreds.begin(), genderPreds.end()));//获得检测结果string gender = genderList[max_index_gender];cout << "Gender: " << gender << endl;//年龄识别ageNet.setInput(blob);vector<float> agePreds = ageNet.forward();// finding maximum indicd in the age_preds vector 找到年龄预测最大下表int max_indice_age = std::distance(agePreds.begin(), max_element(agePreds.begin(), agePreds.end()));string age = ageList[max_indice_age];cout << "Age: " << age << endl;// label 输出标签string label = gender + ", " + age;//在人脸定位图上显示结果cv::putText(frameFace, label, Point(it->at(0), it->at(1) - 15), cv::FONT_HERSHEY_SIMPLEX, 0.9, Scalar(0, 255, 255), 2, cv::LINE_AA);}//保存结果imshow("Frame", frameFace);imwrite("out.jpg", frameFace);}}

python版本：


# Import required modulesimport cv2 as cvimport timeimport argparsedef getFaceBox(net, frame, conf_threshold=0.7):frameOpencvDnn = frame.copy()frameHeight = frameOpencvDnn.shape[0]frameWidth = frameOpencvDnn.shape[1]blob = cv.dnn.blobFromImage(frameOpencvDnn, 1.0, (300, 300), [104, 117, 123], True, False)net.setInput(blob)detections = net.forward()bboxes = []for i in range(detections.shape[2]):confidence = detections[0, 0, i, 2]if confidence > conf_threshold:x1 = int(detections[0, 0, i, 3] * frameWidth)y1 = int(detections[0, 0, i, 4] * frameHeight)x2 = int(detections[0, 0, i, 5] * frameWidth)y2 = int(detections[0, 0, i, 6] * frameHeight)bboxes.append([x1, y1, x2, y2])cv.rectangle(frameOpencvDnn, (x1, y1), (x2, y2), (0, 255, 0), int(round(frameHeight/150)), 8)return frameOpencvDnn, bboxesparser = argparse.ArgumentParser(description='Use this script to run age and gender recognition using OpenCV.')parser.add_argument('--input', help='Path to input image or video file. Skip this argument to capture frames from a camera.')args = parser.parse_args()faceProto = "age_gender/model/opencv_face_detector.pbtxt"faceModel = "age_gender/model/opencv_face_detector_uint8.pb"ageProto = "age_gender/model/age_deploy.prototxt"ageModel = "age_gender/model/age_net.caffemodel"genderProto = "age_gender/model/gender_deploy.prototxt"genderModel = "age_gender/model/gender_net.caffemodel"MODEL_MEAN_VALUES = (78.4263377603, 87.7689143744, 114.895847746)ageList = ['(0-2)', '(4-6)', '(8-12)', '(15-20)', '(25-32)', '(38-43)', '(48-53)', '(60-100)']genderList = ['Male', 'Female']# Load networkageNet = cv.dnn.readNet(ageModel, ageProto)genderNet = cv.dnn.readNet(genderModel, genderProto)faceNet = cv.dnn.readNet(faceModel, faceProto)# Open a video file or an image file or a camera streamcap = cv.VideoCapture(args.input if args.input else 0)padding = 20while cv.waitKey(1) < 0:# Read framet = time.time()hasFrame, frame = cap.read()if not hasFrame:cv.waitKey()breakframeFace, bboxes = getFaceBox(faceNet, frame)if not bboxes:print("No face Detected, Checking next frame")continuefor bbox in bboxes:# print(bbox)face = frame[max(0,bbox[1]-padding):min(bbox[3]+padding,frame.shape[0]-1),max(0,bbox[0]-padding):min(bbox[2]+padding, frame.shape[1]-1)]blob = cv.dnn.blobFromImage(face, 1.0, (227, 227), MODEL_MEAN_VALUES, swapRB=False)genderNet.setInput(blob)genderPreds = genderNet.forward()gender = genderList[genderPreds[0].argmax()]# print("Gender Output : {}".format(genderPreds))print("Gender : {}, conf = {:.3f}".format(gender, genderPreds[0].max()))ageNet.setInput(blob)agePreds = ageNet.forward()age = ageList[agePreds[0].argmax()]print("Age Output : {}".format(agePreds))print("Age : {}, conf = {:.3f}".format(age, agePreds[0].max()))label = "{},{}".format(gender, age)cv.putText(frameFace, label, (bbox[0], bbox[1]-10), cv.FONT_HERSHEY_SIMPLEX, 0.8, (0, 255, 255), 2, cv.LINE_AA)cv.imshow("Age Gender Demo", frameFace)# cv.imwrite("age-gender-out-{}".format(args.input),frameFace)print("time : {:.3f}".format(time.time() - t))