• C# OpenCvSharp Yolov8 Face Landmarks 人脸特征检测

    目录

    介绍

    效果

    模型信息

    项目

    代码

    下载

    介绍

    github地址：https://github.com/derronqi/yolov8-face

    yolov8 face detection with landmark

    效果

    模型信息

    Model Properties
    -------------------------
    description：Ultralytics YOLOv8-lite-t-pose model trained on widerface.yaml
    author：Ultralytics
    kpt_shape：[5, 3]
    task：pose
    license：AGPL-3.0 https://ultralytics.com/license
    version：8.0.85
    stride：32
    batch：1
    imgsz：[640, 640]
    names：{0: 'face'}
    ---------------------------------------------------------------

    Inputs
    -------------------------
    name：images
    tensor：Float[1, 3, 640, 640]
    ---------------------------------------------------------------

    Outputs
    -------------------------
    name：output0
    tensor：Float[1, 80, 80, 80]
    name：884
    tensor：Float[1, 80, 40, 40]
    name：892
    tensor：Float[1, 80, 20, 20]
    ---------------------------------------------------------------

    项目

    代码

    GenerateProposal函数

    public static unsafe void GenerateProposal(int inpHeight, int inpWidth, int reg_max, int num_class, float score_threshold, int feat_h, int feat_w, Mat output, List position_boxes, List confidences, List> landmarks, int imgh, int imgw, float ratioh, float ratiow, int padh, int padw)
    {
        int stride = (int)Math.Ceiling((double)(inpHeight / feat_h));
        int area = feat_h * feat_w;

        float* ptr = (float*)output.DataStart;
        float* ptr_cls = ptr + area * reg_max * 4;
        float* ptr_kp = ptr + area * (reg_max * 4 + num_class);

        for (int i = 0; i < feat_h; i++)
        {
            for (int j = 0; j < feat_w; j++)
            {
                int cls_id = -1;
                float max_conf = -10000;
                int index = i * feat_w + j;
                for (int k = 0; k < num_class; k++)
                {
                    float conf = ptr_cls[k * area + index];
                    if (conf > max_conf)
                    {
                        max_conf = conf;
                        cls_id = k;
                    }
                }
                float box_prob = Common.sigmoid_x(max_conf);
                if (box_prob > score_threshold)
                {
                    float[] pred_ltrb = new float[4];
                    float[] dfl_value = new float[reg_max];
                    float[] dfl_softmax = new float[reg_max];
                    for (int k = 0; k < 4; k++)
                    {
                        for (int n = 0; n < reg_max; n++)
                        {
                            dfl_value[n] = ptr[(k * reg_max + n) * area + index];
                        }
                        Common.softmax_(ref dfl_value, ref dfl_softmax, reg_max);
                        float dis = 0f;
                        for (int n = 0; n < reg_max; n++)
                        {
                            dis += n * dfl_softmax[n];
                        }
                        pred_ltrb[k] = dis * stride;
                    }
                    float cx = (j + 0.5f) * stride;
                    float cy = (i + 0.5f) * stride;
                    float xmin = Math.Max((cx - pred_ltrb[0] - padw) * ratiow, 0f);  ///还原回到原图
                    float ymin = Math.Max((cy - pred_ltrb[1] - padh) * ratioh, 0f);
                    float xmax = Math.Min((cx + pred_ltrb[2] - padw) * ratiow, (float)(imgw - 1));
                    float ymax = Math.Min((cy + pred_ltrb[3] - padh) * ratioh, (float)(imgh - 1));

                    Rect box = new Rect((int)xmin, (int)ymin, (int)(xmax - xmin), (int)(ymax - ymin));

                    position_boxes.Add(box);
                    confidences.Add(box_prob);

                    List kpts = new List();

                    for (int k = 0; k < 5; k++)
                    {
                        float x = ((ptr_kp[(k * 3) * area + index] * 2 + j) * stride - padw) * ratiow;  ///还原回到原图
                        float y = ((ptr_kp[(k * 3 + 1) * area + index] * 2 + i) * stride - padh) * ratioh;
                        kpts.Add(new OpenCvSharp.Point((int)x, (int)y));
                    }
                    landmarks.Add(kpts);
                }
            }
        }
    }

    1. using OpenCvSharp;
    2. using OpenCvSharp.Dnn;
    3. using System;
    4. using System.Collections.Generic;
    5. using System.Drawing;
    6. using System.Linq;
    7. using System.Text;
    8. using System.Windows.Forms;
    9.  
    10. namespace OpenCvSharp_Yolov8_Demo
    11. {
    12.     public partial class frmMain : Form
    13.     {
    14.         public frmMain()
    15.         {
    16.             InitializeComponent();
    17.         }
    18.  
    19.         string fileFilter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";
    20.         string image_path = "";
    21.         string startupPath;
    22.  
    23.         DateTime dt1 = DateTime.Now;
    24.         DateTime dt2 = DateTime.Now;
    25.         string model_path;
    26.         Mat image;
    27.         Mat result_image;
    28.  
    29.         Net opencv_net;
    30.         Mat BN_image;
    31.  
    32.         StringBuilder sb = new StringBuilder();
    33.  
    34.         int reg_max = 16;
    35.         int num_class = 1;
    36.  
    37.         int inpWidth = 640;
    38.         int inpHeight = 640;
    39.  
    40.         float score_threshold = 0.25f;
    41.         float nms_threshold = 0.5f;
    42.  
    43.         private void Form1_Load(object sender, EventArgs e)
    44.         {
    45.             startupPath = System.Windows.Forms.Application.StartupPath;
    46.             model_path = startupPath + "\\yolov8-lite-t.onnx";
    47.             //初始化网络类，读取本地模型
    48.             opencv_net = CvDnn.ReadNetFromOnnx(model_path);
    49.         }
    50.  
    51.         private void button1_Click(object sender, EventArgs e)
    52.         {
    53.             OpenFileDialog ofd = new OpenFileDialog();
    54.             ofd.Filter = fileFilter;
    55.             if (ofd.ShowDialog() != DialogResult.OK) return;
    56.             pictureBox1.Image = null;
    57.             image_path = ofd.FileName;
    58.             pictureBox1.Image = new Bitmap(image_path);
    59.             textBox1.Text = "";
    60.             image = new Mat(image_path);
    61.             pictureBox2.Image = null;
    62.         }
    63.  
    64.         private void button2_Click(object sender, EventArgs e)
    65.         {
    66.             if (image_path == "")
    67.             {
    68.                 return;
    69.             }
    70.             int newh = 0, neww = 0, padh = 0, padw = 0;
    71.             Mat resize_img = Common.ResizeImage(image, inpHeight, inpWidth, ref newh, ref neww, ref padh, ref padw);
    72.             float ratioh = (float)image.Rows / newh, ratiow = (float)image.Cols / neww;
    73.  
    74.             //数据归一化处理
    75.             BN_image = CvDnn.BlobFromImage(resize_img, 1 / 255.0, new OpenCvSharp.Size(inpWidth, inpHeight), new Scalar(0, 0, 0), true, false);
    76.  
    77.             //配置图片输入数据
    78.             opencv_net.SetInput(BN_image);
    79.  
    80.             //模型推理，读取推理结果
    81.             Mat[] outs = new Mat[3] { new Mat(), new Mat(), new Mat() };
    82.             string[] outBlobNames = opencv_net.GetUnconnectedOutLayersNames().ToArray();
    83.  
    84.             dt1 = DateTime.Now;
    85.             opencv_net.Forward(outs, outBlobNames);
    86.             dt2 = DateTime.Now;
    87.  
    88.             List<Rect> position_boxes = new List<Rect>();
    89.             List<float> confidences = new List<float>();
    90.             List<List<OpenCvSharp.Point>> landmarks = new List>();
    92.             Common.GenerateProposal(inpHeight, inpWidth, reg_max, num_class, score_threshold, 40, 40, outs[0], position_boxes, confidences, landmarks, image.Rows, image.Cols, ratioh, ratiow, padh, padw);
    93.             Common.GenerateProposal(inpHeight, inpWidth, reg_max, num_class, score_threshold, 20, 20, outs[1], position_boxes, confidences, landmarks, image.Rows, image.Cols, ratioh, ratiow, padh, padw);
    94.             Common.GenerateProposal(inpHeight, inpWidth, reg_max, num_class, score_threshold, 80, 80, outs[2], position_boxes, confidences, landmarks, image.Rows, image.Cols, ratioh, ratiow, padh, padw);
    95.  
    96.             //NMS非极大值抑制
    97.             int[] indexes = new int[position_boxes.Count];
    98.             CvDnn.NMSBoxes(position_boxes, confidences, score_threshold, nms_threshold, out indexes);
    99.  
    100.             List<Rect> re_result = new List<Rect>();
    101.             List<List<OpenCvSharp.Point>> re_landmarks = new List>();
    102.             List<float> re_confidences = new List<float>();
    103.  
    104.             for (int i = 0; i < indexes.Length; i++)
    105.             {
    106.                 int index = indexes[i];
    107.                 re_result.Add(position_boxes[index]);
    108.                 re_landmarks.Add(landmarks[index]);
    109.                 re_confidences.Add(confidences[index]);
    110.             }
    111.  
    112.             if (re_result.Count > 0)
    113.             {
    114.                 sb.Clear();
    115.                 sb.AppendLine("推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms");
    116.                 sb.AppendLine("--------------------------");
    117.  
    118.                 //将识别结果绘制到图片上
    119.                 result_image = image.Clone();
    120.  
    121.                 for (int i = 0; i < re_result.Count; i++)
    122.                 {
    123.                     Cv2.Rectangle(result_image, re_result[i], new Scalar(0, 0, 255), 2, LineTypes.Link8);
    124.  
    125.                     Cv2.PutText(result_image, "face-" + re_confidences[i].ToString("0.00"),
    126.                         new OpenCvSharp.Point(re_result[i].X, re_result[i].Y - 10),
    127.                         HersheyFonts.HersheySimplex, 1, new Scalar(0, 0, 255), 2);
    128.  
    129.                     foreach (var item in re_landmarks[i])
    130.                     {
    131.                         Cv2.Circle(result_image, item, 4, new Scalar(0, 255, 0), -1);
    132.                     }
    133.  
    134.                     sb.AppendLine(string.Format("{0}：{1},({2},{3},{4},{5})"
    135.                         , "face"
    136.                         , re_confidences[i].ToString("0.00")
    137.                         , re_result[i].TopLeft.X
    138.                         , re_result[i].TopLeft.Y
    139.                         , re_result[i].BottomRight.X
    140.                         , re_result[i].BottomRight.Y
    141.                         ));
    142.                 }
    143.  
    144.                 pictureBox2.Image = new Bitmap(result_image.ToMemoryStream());
    145.                 textBox1.Text = sb.ToString();
    146.  
    147.             }
    148.             else
    149.             {
    150.                 textBox1.Text = "无信息";
    151.             }
    152.  
    153.         }
    154.  
    155.     }
    156. }

    下载

    exe可执行程序包免费下载

    源码下载

  • 相关阅读:
    ATFX汇市：美国5月PCE数据来袭，EURUSD或迎剧烈波动
    java毕业生设计学习类视频网计算机源码+系统+mysql+调试部署+lw
    Java核心技术卷Ⅰ-第一章Java程序设计概述
    【Redis专题】RedisCluster集群运维与核心原理剖析
    Java：什么是Java框架?
    常见外贸英语缩写词
    如何在 ASP.NET Core 中使用 Quartz.NET
    AD9680配置与数据还原说明
    Codeforces Round 953 (Div. 2) A - C 题解
    Leetcode754. 到达终点数字 --数论+思维
  • 原文地址：https://blog.csdn.net/lw112190/article/details/134077292