object detection

opencv-yolo/yolo.hpp

+#ifndef _MY_YOLO_HPP_
+#define _MY_YOLO_HPP_
+
+#include <fstream>
+#include <sstream>
+#include <iostream>
+#include <tuple>
+#include "fs.h"
+#include "spdlog/spdlog.h"
+
+#ifdef _MY_HEADERS_
+#include <opencv2/core/types_c.h>
+#include <opencv2/dnn.hpp>
+#include <opencv2/imgproc.hpp>
+#include <opencv2/highgui.hpp>
+#else
+#include <opencv2/core/types_c.h>
+#include <opencv2/opencv.hpp>
+#endif
+
+using namespace cv;
+using namespace dnn;
+using namespace std;
+
+class YoloDectect {
+public:
+    unsigned long numFrameProcessed = 0;
+private:
+    // Initialize the parameters
+    const string selfId = "YoloDetector";
+    float confThreshold = 0.5; // Confidence threshold
+    float nmsThreshold = 0.4;  // Non-maximum suppression threshold
+    int inpWidth = 416;  // Width of network's input image
+    int inpHeight = 416; // Height of network's input image
+    vector<string> classes;
+    Net net;
+    Mat blob;
+    VideoCapture cap;
+    VideoWriter video;
+    bool bOutputIsImg = false;
+    string outFileBase;
+    bool cmdStop = false;
+    unsigned int wrapNum = 0;
+    unsigned int numLogSkip = 0;
+
+    // Get the names of the output layers
+    vector<String> getOutputsNames(const Net& net)
+    {
+        static vector<String> names;
+        if (names.empty()) {
+            //Get the indices of the output layers, i.e. the layers with unconnected outputs
+            vector<int> outLayers = net.getUnconnectedOutLayers();
+
+            //get the names of all the layers in the network
+            vector<String> layersNames = net.getLayerNames();
+
+            // Get the names of the output layers in names
+            names.resize(outLayers.size());
+            for (size_t i = 0; i < outLayers.size(); ++i)
+                names[i] = layersNames[outLayers[i] - 1];
+        }
+        return names;
+    }
+
+    // draw the predicted bounding box
+    void drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
+    {
+        // draw a rectangle displaying the bounding box
+        rectangle(frame, Point(left, top), Point(right, bottom), Scalar(255, 178, 50), 3);
+
+        //get the label for the class name and its confidence
+        string label = format("%.2f", conf);
+        if (!classes.empty()) {
+            CV_Assert(classId < (int)classes.size());
+            label = classes[classId] + ":" + label;
+        }
+
+        // display the label at the top of the bounding box
+        int baseLine;
+        Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
+        top = max(top, labelSize.height);
+        rectangle(frame, Point(left, top - round(1.5*labelSize.height)), Point(left + round(1.5*labelSize.width), top + baseLine), Scalar(255, 255, 255), FILLED);
+        putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.75, Scalar(0,0,0),1);
+    }
+
+    // post process
+    vector<tuple<string, double, Rect>> postprocess(Mat& frame, const vector<Mat>& outs, bool bModify = false)
+    {
+        vector<int> classIds;
+        vector<float> confidences;
+        vector<Rect> boxes;
+
+        for (size_t i = 0; i < outs.size(); ++i) {
+            // Scan through all the bounding boxes output from the network and keep only the
+            // ones with high confidence scores. Assign the box's class label as the class
+            // with the highest score for the box.
+            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;
+                // Get the value and location of the maximum score
+                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));
+                }
+            }
+        }
+
+        // Perform non maximum suppression to eliminate redundant overlapping boxes with lower confidences
+        vector<int> indices;
+        NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, indices);
+        vector<tuple<string, double, Rect>> ret;
+        for (size_t i = 0; i < indices.size(); ++i) {
+            int idx = indices[i];
+            Rect box = boxes[idx];
+            ret.push_back(tuple<string, double, Rect>(classes[classIds[idx]], confidences[idx], box));
+            if(bModify)
+                drawPred(classIds[idx], confidences[idx], box.x, box.y, box.x + box.width, box.y + box.height, frame);
+        }
+
+        return ret;
+    }
+
+    //
+protected:
+
+    //
+public:
+    typedef int (*callback)(vector<tuple<string, double, Rect>>&, Mat);
+    YoloDectect(string path = ".", unsigned int _wrapNum = 10, unsigned int _numLogSkip = 380)
+    {
+        if(path.empty()) {
+            path = ".";
+        }
+
+        wrapNum = _wrapNum;
+        numLogSkip = _numLogSkip;
+        
+        // Load names of classes
+        string classesFile = path + "/coco.names";
+        // Give the configuration and weight files for the model
+        String modCfg = path + "/yolov3-tiny.cfg";
+        String modWeights = path + "/yolov3-tiny.weights";
+
+        if(!fs::exists(classesFile) || !fs::exists(modCfg) || !fs::exists(modWeights)) {
+            spdlog::error("{} failed to load configration files", selfId);
+            exit(1);
+        }
+
+        ifstream ifs(classesFile.c_str());
+        string line;
+        while (getline(ifs, line)) {
+            classes.push_back(line);
+        }
+
+        // Load the network
+        net = readNetFromDarknet(modCfg, modWeights);
+        net.setPreferableBackend(DNN_BACKEND_OPENCV);
+        net.setPreferableTarget(DNN_TARGET_CPU);
+        spdlog::info("{} inited", selfId);
+    }
+
+    vector<tuple<string, double, Rect>> process(Mat &inFrame, Mat* pOutFrame, bool bModify = false)
+    {
+        if(inFrame.empty()) {
+            return vector<tuple<string, double, Rect>>();
+        }
+
+        // Create a 4D blob from a frame.
+        blobFromImage(inFrame, blob, 1/255.0, cvSize(inpWidth, inpHeight), Scalar(0,0,0), true, false);
+
+        //Sets the input to the network
+        net.setInput(blob);
+
+        // Runs the forward pass to get output of the output layers
+        vector<Mat> outs;
+        net.forward(outs, getOutputsNames(net));
+
+        // Remove the bounding boxes with low confidence
+        auto ret = postprocess(inFrame, outs, bModify);
+
+        // The function getPerfProfile returns the overall time for inference(t) and the timings for each of the layers(in layersTimes)
+        vector<double> layersTimes;
+        if(numLogSkip == 0 || numFrameProcessed % numLogSkip == 0) {
+            double freq = getTickFrequency() / 1000;
+            double t = net.getPerfProfile(layersTimes) / freq;
+            spdlog::info("{} infer time: {} ms", selfId, t);
+        }
+        if(pOutFrame != nullptr){
+            inFrame.convertTo(*pOutFrame, CV_8U);
+        }
+        
+        numFrameProcessed++;
+        return ret;
+    }
+
+    int process(string inVideoUri, string outFile = "processed.jpg", bool bHumanExit = false, callback cb = nullptr)
+    {
+        if(inVideoUri.empty()) {
+            inVideoUri = "0";
+        }
+
+        if(!cap.open(inVideoUri)) {
+            spdlog::error("{} failed to open input video {}", selfId, inVideoUri);
+            return -1;
+        }
+
+        ghc::filesystem::path p(outFile);
+        auto dir = p.parent_path();
+
+        if((outFile.substr(outFile.find_last_of(".") + 1) == "jpg")) {
+            bOutputIsImg = true;
+            outFileBase = string(dir / p.stem());
+            spdlog::info("{} outFileBase {}", selfId, outFileBase);
+        }
+        else {
+            bOutputIsImg = false;
+            if(!video.open(outFile, VideoWriter::fourcc('M','J','P','G'), 28, Size(cap.get(CAP_PROP_FRAME_WIDTH), cap.get(CAP_PROP_FRAME_HEIGHT)))) {
+                spdlog::error("{} failed to open output video {}", selfId, outFile);
+                return -1;
+            }
+        }
+
+        spdlog::info("{} try to process video {} to {}", selfId, inVideoUri, outFile);
+
+        unsigned long frameCnt = 0;
+        unsigned long detCnt = 0, skipCnt = 0;
+        Mat frame, outFrame;
+        while (true) {
+            // get frame from the video
+            if(cmdStop) {
+                break;
+            }
+
+            if(!cap.read(frame)) {
+                spdlog::info("{} done reading frame from {}", selfId, inVideoUri);
+                break;
+            }
+
+            frameCnt++;
+            if(frameCnt %100 == 0)
+                spdlog::info("framecnt {}", frameCnt);
+            
+            if(frameCnt % 30  != 0 ){
+                continue;
+            }
+
+            // Stop the program if reached end of video
+            if (frame.empty()) {
+                continue;
+            }
+
+            vector<tuple<string, double, Rect>> ret = process(frame, &outFrame, true);
+            if(cb == nullptr) {
+                if(ret.size() == 0 && bOutputIsImg) {
+                    // no detection
+                    if(numLogSkip == 0|| skipCnt % numLogSkip == 0) {
+                        spdlog::info("{} no valid object detected skipped frame count {}", selfId, skipCnt);
+                    }
+                    skipCnt++;    
+                    continue;
+                }
+                if(bHumanExit){
+                    for(auto &[s, c, r]:ret) {
+                        if (s == "person"){
+                            string ofname = outFileBase + "_person.jpg";
+                            imwrite(ofname, outFrame); 
+                            spdlog::info("found human {} x: {}, y: {}, w: {}, h: {}", c, r.x, r.y, r.width, r.height);
+                            cmdStop = true;
+                            break;
+                        }
+                    }
+                }
+
+                if (bOutputIsImg) {
+                    if(wrapNum > 0) {
+                        detCnt = detCnt % wrapNum;
+                    }
+
+                    string ofname = outFileBase + to_string(detCnt) + ".jpg";
+                    imwrite(ofname, outFrame);
+                    detCnt++;
+                }
+                else {
+                    video.write(outFrame);
+                }
+            }else{
+                cb(ret, outFrame);
+            }
+        }
+
+        cap.release();
+        if(!bOutputIsImg) video.release();
+
+        return 0;
+    }
+};
+
+#endif
\ No newline at end of file