C# OpenCV EmguCv mobilenet_v3 对象检测

本文主要是介绍C# OpenCV EmguCv mobilenet_v3 对象检测，对大家解决编程问题具有一定的参考价值，需要的程序猿们随着小编来一起学习吧！

链接：https://pan.baidu.com/s/1I9N3lRe_t9C24IKJRcv89Q?pwd=1212

由于OpenCVSharp缺少相关api，这次使用EmguCv，这个demo网上都是python版本的，抄写为C#

使用ssd_mobilenet_v3_large_coco_2020_01_14.pbtxt和frozen_inference_graph.pb模型

public class EasyObjectDetection
{
    public static void Run()
    {
        string modelPathBase = @"D:\SourceCode\Models\";
        string classesfile = modelPathBase + "coco.names";
        string configPath = modelPathBase + "ssd_mobilenet_v3_large_coco_2020_01_14.pbtxt";
        string weightsPath = modelPathBase + "frozen_inference_graph.pb";
        var thres = 0.5f;
        var cap = new VideoCapture(modelPathBase + "Road_traffic_video2.mp4");
        string[] classes = File.ReadAllLines(classesfile);
        var net = new Emgu.CV.Dnn.DetectionModel(weightsPath, configPath);
        net.SetInputSize(new Size(320, 320));
        net.SetInputScale(1.0 / 127.5);
        net.SetInputMean(new Emgu.CV.Structure.MCvScalar(127.5, 127.5, 127.5));
        net.SetInputSwapRB(true);
        Mat frame = new Mat();
        while (true)
        {
            var isSuccess = cap.Read(frame);
            var classIds = new VectorOfInt();
            var confs = new VectorOfFloat();
            var bbox = new VectorOfRect();
            var indices = new VectorOfInt();
            net.Detect(frame, classIds, confs, bbox, confThreshold: 0.5f);
            DnnInvoke.NMSBoxes(bbox, confs, thres, 0.2f, indices);
            for (int i = 0; i < indices.Size; i++)
            {
                var confidence = confs[i].ToString();
                for (int j = 0; j < classIds.Size; j++)
                {
                    var classId = classIds[j];
                    //var box = bbox[j];
                    Rectangle box = bbox[indices[i]];
                    CvInvoke.Rectangle(frame, box, new Emgu.CV.Structure.MCvScalar(255, 0, 255), 2);
                    CvInvoke.PutText(frame, classes[classIds[i] - 1].ToUpper(), new Point(box.X + 2, box.Y + 2),
                         Emgu.CV.CvEnum.FontFace.HersheyPlain, 0.7, new Emgu.CV.Structure.MCvScalar(200, 200, 200), 1);
                }
            }
            CvInvoke.Imshow("output", frame);
            int c = CvInvoke.WaitKey(5);
            if (c == 27)
            { // ESC退出
                break;
            }
        }
    }
}

这篇关于C# OpenCV EmguCv mobilenet_v3 对象检测的文章就介绍到这儿，希望我们推荐的文章对大家有所帮助，也希望大家多多支持为之网！