人脸识别引擎SeetaFaceEngine中Identification模块使用的测试代码

人脸识别引擎SeetaFaceEngine中Identification模块用于比较两幅人脸图像的相似度，以下是测试代码：

int test_recognize()
{const std::string path_images{ "E:/GitCode/Face_Test/testdata/recognization/" };seeta::FaceDetection detector("E:/GitCode/Face_Test/src/SeetaFaceEngine/FaceDetection/model/seeta_fd_frontal_v1.0.bin");seeta::FaceAlignment alignment("E:/GitCode/Face_Test/src/SeetaFaceEngine/FaceAlignment/model/seeta_fa_v1.1.bin");seeta::FaceIdentification face_recognizer("E:/GitCode/Face_Test/src/SeetaFaceEngine/FaceIdentification/model/seeta_fr_v1.0.bin");detector.SetMinFaceSize(20);detector.SetMaxFaceSize(200);detector.SetScoreThresh(2.f);detector.SetImagePyramidScaleFactor(0.8f);detector.SetWindowStep(4, 4);std::vector<std::vector<seeta::FacialLandmark>> landmards;// detect and alignmentfor (int i = 0; i < 20; i++) {std::string image = path_images + std::to_string(i) + ".jpg";//fprintf(stderr, "start process image: %s\n", image.c_str());cv::Mat src_ = cv::imread(image, 1);if (src_.empty()) {fprintf(stderr, "read image error: %s\n", image.c_str());continue;}cv::Mat src;cv::cvtColor(src_, src, CV_BGR2GRAY);seeta::ImageData img_data;img_data.data = src.data;img_data.width = src.cols;img_data.height = src.rows;img_data.num_channels = 1;std::vector<seeta::FaceInfo> faces = detector.Detect(img_data);if (faces.size() == 0) {fprintf(stderr, "%s don't detect face\n", image.c_str());continue;}// Detect 5 facial landmarks: two eye centers, nose tip and two mouth cornersstd::vector<seeta::FacialLandmark> landmard(5);alignment.PointDetectLandmarks(img_data, faces[0], &landmard[0]);landmards.push_back(landmard);cv::rectangle(src_, cv::Rect(faces[0].bbox.x, faces[0].bbox.y,faces[0].bbox.width, faces[0].bbox.height), cv::Scalar(0, 255, 0), 2);for (auto point : landmard) {cv::circle(src_, cv::Point(point.x, point.y), 2, cv::Scalar(0, 0, 255), 2);}std::string save_result = path_images + "_" + std::to_string(i) + ".jpg";cv::imwrite(save_result, src_);}int width = 200;int height = 200;cv::Mat dst(height * 5, width * 4, CV_8UC3);for (int i = 0; i < 20; i++) {std::string input_image = path_images + "_" + std::to_string(i) + ".jpg";cv::Mat src = cv::imread(input_image, 1);if (src.empty()) {fprintf(stderr, "read image error: %s\n", input_image.c_str());return -1;}cv::resize(src, src, cv::Size(width, height), 0, 0, 4);int x = (i * width) % (width * 4);int y = (i / 4) * height;cv::Mat part = dst(cv::Rect(x, y, width, height));src.copyTo(part);}std::string output_image = path_images + "result_alignment.png";cv::imwrite(output_image, dst);// crop imagefor (int i = 0; i < 20; i++) {std::string image = path_images + std::to_string(i) + ".jpg";//fprintf(stderr, "start process image: %s\n", image.c_str());cv::Mat src_img = cv::imread(image, 1);if (src_img.data == nullptr) {fprintf(stderr, "Load image error: %s\n", image.c_str());return -1;}if (face_recognizer.crop_channels() != src_img.channels()) {fprintf(stderr, "channels dismatch: %d, %d\n", face_recognizer.crop_channels(), src_img.channels());return -1;}// ImageData store data of an image without memory alignment.seeta::ImageData src_img_data(src_img.cols, src_img.rows, src_img.channels());src_img_data.data = src_img.data;// Create a image to store crop face.cv::Mat dst_img(face_recognizer.crop_height(), face_recognizer.crop_width(), CV_8UC(face_recognizer.crop_channels()));seeta::ImageData dst_img_data(dst_img.cols, dst_img.rows, dst_img.channels());dst_img_data.data = dst_img.data;// Crop Faceface_recognizer.CropFace(src_img_data, &landmards[i][0], dst_img_data);std::string save_image_name = path_images + "crop_" + std::to_string(i) + ".jpg";cv::imwrite(save_image_name, dst_img);}dst = cv::Mat(height * 5, width * 4, CV_8UC3);for (int i = 0; i < 20; i++) {std::string input_image = path_images + "crop_" + std::to_string(i) + ".jpg";cv::Mat src_img = cv::imread(input_image, 1);if (src_img.empty()) {fprintf(stderr, "read image error: %s\n", input_image.c_str());return -1;}cv::resize(src_img, src_img, cv::Size(width, height), 0, 0, 4);int x = (i * width) % (width * 4);int y = (i / 4) * height;cv::Mat part = dst(cv::Rect(x, y, width, height));src_img.copyTo(part);}output_image = path_images + "result_crop.png";cv::imwrite(output_image, dst);// extract featureint feat_size = face_recognizer.feature_size();if (feat_size != 2048) {fprintf(stderr, "feature size mismatch: %d\n", feat_size);return -1;}float* feat_sdk = new float[feat_size * 20];for (int i = 0; i < 20; i++) {std::string input_image = path_images + "crop_" + std::to_string(i) + ".jpg";cv::Mat src_img = cv::imread(input_image, 1);if (src_img.empty()) {fprintf(stderr, "read image error: %s\n", input_image.c_str());return -1;}cv::resize(src_img, src_img, cv::Size(face_recognizer.crop_height(), face_recognizer.crop_width()));// ImageData store data of an image without memory alignment.seeta::ImageData src_img_data(src_img.cols, src_img.rows, src_img.channels());src_img_data.data = src_img.data;// Extract featureface_recognizer.ExtractFeature(src_img_data, feat_sdk + i * feat_size);}float* feat1 = feat_sdk;// varify(recognize)for (int i = 1; i < 20; i++) {std::string image = std::to_string(i) + ".jpg";float* feat_other = feat_sdk + i * feat_size;// Caculate similarityfloat sim = face_recognizer.CalcSimilarity(feat1, feat_other);fprintf(stdout, "0.jpg -- %s similarity: %f\n", image.c_str(), sim);}delete[] feat_sdk;return 0;
}

从网上找了20张图像，前19张为周星驰，最后一张为汤唯，用于测试此模块，测试结果如下：

detect/alignment结果如下：

crop结果如下：

取上图中最左上图为标准图，与其它19幅图作验证，测试结果如下：

GitHub：https://github.com/fengbingchun/Face_Test

人脸识别引擎SeetaFaceEngine中Identification模块使用的测试代码相关推荐

人脸识别引擎SeetaFaceEngine中Alignment模块使用的测试代码
人脸识别引擎SeetaFaceEngine中Alignment模块用于检测人脸关键点,包括5个点,两个眼的中心.鼻尖.两个嘴角,以下是测试代码: int test_alignment() {std:: ...
人脸识别引擎SeetaFaceEngine中Detection模块使用的测试代码
人脸识别引擎SeetaFaceEngine中Detection模块用于人脸检测,以下是测试代码: int test_detection() {std::vector<std::string> ...
人脸识别引擎SeetaFaceEngine简介及在windows7 vs2013下的编译
SeetaFaceEngine是开源的C++人脸识别引擎,无需第三方库,它是由中科院计算所山世光老师团队研发.它的License是BSD-2. SeetaFaceEngine库包括三个模块:人脸检测( ...
python开源的人脸识别库_什么是 SeetaFace 开源人脸识别引擎
区分不同的人是很多智能系统的必备能力.为实现此目的,一种可能的技术手段是通过对人脸的光学成像来感知人.识别人,即所谓的人脸识别技术.经过几十年的研发积累,特别是近年来深度学习技术的涌现,人脸识别取得了 ...
SeetaFace开源人脸识别引擎
开源网址目前,SeetaFace开源人脸识别引擎已全部发布在Github上供国内外同行和工业界使用,项目网址为:https://github.com/seetaface/SeetaFaceEngin ...
【SeetaFace】中科院山世光老师开源的Seetaface人脸识别引擎测试
Seetaface发布了第二个版本--Seetaface2: https://github.com/seetaface/SeetaFaceEngine2 SeetaFace2 Engine conta ...
SeetaFace开源人脸识别引擎介绍
深度学习大讲堂致力于推送人工智能,深度学习方面的最新技术,产品以及活动.请关注我们的知乎专栏! 区分不同的人是很多智能系统的必备能力.为实现此目的,一种可能的技术手段是通过对人脸的光学成像来感知人.识 ...
C/C++人脸识别引擎 SeetaFace2打造“刷脸时代”
SeetaFace2 人脸识别引擎包括了搭建一套全自动人脸识别系统所需的三个核心模块,即:人脸检测模块 FaceDetector.面部关键点定位模块 FaceLandmarker 以及人脸特征提取与比 ...
C# 实现人脸识别一（运用虹软人脸识别引擎）
知识背景: 下载虹软人脸识别引擎下载地址: http://www.arcsoft.com.cn/ai/arcface.html 目前虹软人脸识别引擎有3个平台,其中Windows与iOS是基于C++ ...

人脸识别引擎SeetaFaceEngine中Identification模块使用的测试代码

人脸识别引擎SeetaFaceEngine中Identification模块使用的测试代码相关推荐

最新文章

热门文章