opencv dnn模塊 示例(4) 目標檢測 object_detection (3) caffe SSD

一、opencv的示例模型文件

opencv的dnn模塊讀取models.yml文件中包含的目標檢測模型有5種，這裏實例Caffe implementation of SSD model。
from https://github.com/chuanqi305/MobileNet-SSD
ssd_caffe:
model: “MobileNetSSD_deploy.caffemodel”
config: “MobileNetSSD_deploy.prototxt”
mean: [127.5, 127.5, 127.5]
scale: 0.007843
width: 300
height: 300
rgb: false
classes: “object_detection_classes_pascal_voc.txt”
sample: “object_detection”

二、示例代碼

整體代碼和前面相同，僅配置不同

#include <fstream>
#include <sstream>

#include <opencv2/dnn.hpp>
#include <opencv2/imgproc.hpp>
#include <opencv2/highgui.hpp>

using namespace cv;
using namespace dnn;

float confThreshold, nmsThreshold;
std::vector<std::string> classes;

void postprocess(Mat& frame, const std::vector<Mat>& out, Net& net);

void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame);

int main(int argc, char** argv)
{
	// 根據選擇的檢測模型文件進行配置 
	confThreshold = 0.5;
	nmsThreshold = 0.4;

	float scale = 0.007843;
	Scalar mean = { 127.5, 127.5, 127.5 };
	bool swapRB =  false;
	int inpWidth = 300;
	int inpHeight = 300;
	
	String modelPath = "../../data/testdata/dnn/MobileNetSSD_deploy.caffemodel";
	String configPath = "../../data/testdata/dnn/MobileNetSSD_deploy.prototxt";
	String framework = "";

	int backendId = cv::dnn::DNN_BACKEND_OPENCV;
	int targetId = cv::dnn::DNN_TARGET_OPENCL; // cv::dnn::DNN_TARGET_CPU

	String classesFile = R"(../../data/dnn/object_detection_classes_pascal_voc.txt)";

	// Open file with classes names.
	if (!classesFile.empty()) {
		const std::string& file = classesFile;
		std::ifstream ifs(file.c_str());
		if (!ifs.is_open())
			CV_Error(Error::StsError, "File " + file + " not found");
		std::string line;
		while (std::getline(ifs, line)) {
			classes.push_back(line);
		}
	}


	// Load a model.
	Net net = readNet(modelPath, configPath, framework);
	net.setPreferableBackend(backendId);
	net.setPreferableTarget(targetId);


	std::vector<String> outNames = net.getUnconnectedOutLayersNames();

	// Create a window
	static const std::string kWinName = "Deep learning object detection in OpenCV";

	// Open a video file or an image file or a camera stream.
	VideoCapture cap;
	cap.open(0);

	// Process frames.
	Mat frame, blob;
	while (waitKey(1) < 0) {
		cap >> frame;
		if (frame.empty()) {
			waitKey();
			break;
		}

		// Create a 4D blob from a frame.
		Size inpSize(inpWidth > 0 ? inpWidth : frame.cols,
			inpHeight > 0 ? inpHeight : frame.rows);
		blobFromImage(frame, blob, scale, inpSize, mean, swapRB, false);

		// Run a model.
		net.setInput(blob);
		if (net.getLayer(0)->outputNameToIndex("im_info") != -1)  // Faster-RCNN or R-FCN
		{
			resize(frame, frame, inpSize);
			Mat imInfo = (Mat_<float>(1, 3) << inpSize.height, inpSize.width, 1.6f);
			net.setInput(imInfo, "im_info");
		}

		std::vector<Mat> outs;
		net.forward(outs, outNames);

		postprocess(frame, outs, net);

		// Put efficiency information.
		std::vector<double> layersTimes;
		double freq = getTickFrequency() / 1000;
		double t = net.getPerfProfile(layersTimes) / freq;
		std::string label = format("Inference time: %.2f ms", t);
		putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 255, 0));

		imshow(kWinName, frame);
	}
	return 0;
}

void postprocess(Mat& frame, const std::vector<Mat>& outs, Net& net)
{
	static std::vector<int> outLayers = net.getUnconnectedOutLayers();
	static std::string outLayerType = net.getLayer(outLayers[0])->type;

	std::vector<int> classIds;
	std::vector<float> confidences;
	std::vector<Rect> boxes;
	if (net.getLayer(0)->outputNameToIndex("im_info") != -1)  // Faster-RCNN or R-FCN
	{
		// Network produces output blob with a shape 1x1xNx7 where N is a number of
		// detections and an every detection is a vector of values
		// [batchId, classId, confidence, left, top, right, bottom]
		CV_Assert(outs.size() == 1);
		float* data = (float*)outs[0].data;
		for (size_t i = 0; i < outs[0].total(); i += 7) {
			float confidence = data[i + 2];
			if (confidence > confThreshold) {
				int left = (int)data[i + 3];
				int top = (int)data[i + 4];
				int right = (int)data[i + 5];
				int bottom = (int)data[i + 6];
				int width = right - left + 1;
				int height = bottom - top + 1;
				classIds.push_back((int)(data[i + 1]) - 1);  // Skip 0th background class id.
				boxes.push_back(Rect(left, top, width, height));
				confidences.push_back(confidence);
			}
		}
	}
	else if (outLayerType == "DetectionOutput") {
		// Network produces output blob with a shape 1x1xNx7 where N is a number of
		// detections and an every detection is a vector of values
		// [batchId, classId, confidence, left, top, right, bottom]
		CV_Assert(outs.size() == 1);
		float* data = (float*)outs[0].data;
		for (size_t i = 0; i < outs[0].total(); i += 7) {
			float confidence = data[i + 2];
			if (confidence > confThreshold) {
				int left = (int)(data[i + 3] * frame.cols);
				int top = (int)(data[i + 4] * frame.rows);
				int right = (int)(data[i + 5] * frame.cols);
				int bottom = (int)(data[i + 6] * frame.rows);
				int width = right - left + 1;
				int height = bottom - top + 1;
				classIds.push_back((int)(data[i + 1]) - 1);  // Skip 0th background class id.
				boxes.push_back(Rect(left, top, width, height));
				confidences.push_back(confidence);
			}
		}
	}
	else if (outLayerType == "Region") {
		for (size_t i = 0; i < outs.size(); ++i) {
			// Network produces output blob with a shape NxC where N is a number of
			// detected objects and C is a number of classes + 4 where the first 4
			// numbers are [center_x, center_y, width, height]
			float* data = (float*)outs[i].data;
			for (int j = 0; j < outs[i].rows; ++j, data += outs[i].cols) {
				Mat scores = outs[i].row(j).colRange(5, outs[i].cols);
				Point classIdPoint;
				double confidence;
				minMaxLoc(scores, 0, &confidence, 0, &classIdPoint);
				if (confidence > confThreshold) {
					int centerX = (int)(data[0] * frame.cols);
					int centerY = (int)(data[1] * frame.rows);
					int width = (int)(data[2] * frame.cols);
					int height = (int)(data[3] * frame.rows);
					int left = centerX - width / 2;
					int top = centerY - height / 2;

					classIds.push_back(classIdPoint.x);
					confidences.push_back((float)confidence);
					boxes.push_back(Rect(left, top, width, height));
				}
			}
		}
	}
	else
		CV_Error(Error::StsNotImplemented, "Unknown output layer type: " + outLayerType);

	std::vector<int> indices;
	NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
	for (size_t i = 0; i < indices.size(); ++i) {
		int idx = indices[i];
		Rect box = boxes[idx];
		drawPred(classIds[idx], confidences[idx], box.x, box.y,
			box.x + box.width, box.y + box.height, frame);
	}
}

void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
{
	rectangle(frame, Point(left, top), Point(right, bottom), Scalar(0, 255, 0));

	std::string label = format("%.2f", conf);
	if (!classes.empty()) {
		CV_Assert(classId < (int)classes.size());
		label = classes[classId] + ": " + label;
	}

	int baseLine;
	Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);

	top = max(top, labelSize.height);
	rectangle(frame, Point(left, top - labelSize.height),
		Point(left + labelSize.width, top + baseLine), Scalar::all(255), FILLED);
	putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.5, Scalar());
}

三、演示

相比於yolo v3， 實際檢測效果要差一些。
使用opencl，每幀~120ms，cpu佔用40%， mem 170M， GPU佔用 ~1.5%；
使用cpu，每幀 ~40ms左右， cpu佔用70%， mem 163M， GPU佔用5%；
這裏有點疑問：使用cpu時 gpu佔用反而高了？？？部分GPU用於圖像實時顯示。