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 alignment
	for (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 corners
		std::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 image
	for (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 Face
		face_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 feature
	int 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 feature
		face_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 similarity
		float 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