C# OpenCvSharp DNN FreeYOLO 人脸检测&人脸图像质量评估

目录

效果

模型信息

yolo_free_huge_widerface_192x320.onnx

face-quality-assessment.onnx

项目

代码

frmMain.cs

FreeYoloFace

FaceQualityAssessment.cs

下载


C# OpenCvSharp DNN FreeYOLO 人脸检测&人脸图像质量评估

效果

模型信息

yolo_free_huge_widerface_192x320.onnx

Inputs


name:input

tensor:Float[1, 3, 192, 320]


Outputs


name:output

tensor:Float[1, 1260, 6]


face-quality-assessment.onnx

Inputs


name:input

tensor:Float[1, 3, 112, 112]


Outputs


name:quality

tensor:Float[1, 10]


项目

代码

frmMain.cs

using OpenCvSharp;

using System;

using System.Collections.Generic;

using System.Drawing;

using System.Text;

using System.Windows.Forms;

namespace OpenCvSharp_DNN_Demo

{

public partial class frmMain : Form

{

public frmMain()

{

InitializeComponent();

}

string fileFilter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";

string image_path = "";

DateTime dt1 = DateTime.Now;

DateTime dt2 = DateTime.Now;

StringBuilder sb = new StringBuilder();

Mat image;

Mat result_image;

FaceQualityAssessment fqa = new FaceQualityAssessment("model/face-quality-assessment.onnx");

FreeYoloFace face = new FreeYoloFace("model/yolo_free_huge_widerface_192x320.onnx");

private void button1_Click(object sender, EventArgs e)

{

OpenFileDialog ofd = new OpenFileDialog();

ofd.Filter = fileFilter;

if (ofd.ShowDialog() != DialogResult.OK) return;

pictureBox1.Image = null;

pictureBox2.Image = null;

textBox1.Text = "";

image_path = ofd.FileName;

pictureBox1.Image = new Bitmap(image_path);

image = new Mat(image_path);

}

private void Form1_Load(object sender, EventArgs e)

{

image_path = "test_img/1.jpg";

pictureBox1.Image = new Bitmap(image_path);

}

private unsafe void button2_Click(object sender, EventArgs e)

{

if (image_path == "")

{

return;

}

textBox1.Text = "检测中,请稍等......";

if (pictureBox2.Image != null)

{

pictureBox2.Image.Dispose();

}

pictureBox2.Image = null;

sb.Clear();

Application.DoEvents();

image = new Mat(image_path);

dt1 = DateTime.Now;

List<Face> ltFace = face.Detect(image);

dt2 = DateTime.Now;

if (ltFace.Count > 0)

{

sb.AppendLine("推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms");

result_image = image.Clone();

foreach (var item in ltFace)

{

Mat crop_img = new Mat(image, item.rect);

float fqa_prob_mean = fqa.Detect(crop_img);

crop_img.Dispose();

Cv2.Rectangle(result_image, new OpenCvSharp.Point(item.rect.X, item.rect.Y), new OpenCvSharp.Point(item.rect.X + item.rect.Width, item.rect.Y + item.rect.Height), new Scalar(0, 0, 255), 2);

string label = "prob:" + item.prob.ToString("0.00") + " fqa_score:" + fqa_prob_mean.ToString("0.00");

sb.AppendLine(label);

Cv2.PutText(result_image, label, new OpenCvSharp.Point(item.rect.X, item.rect.Y - 5), HersheyFonts.HersheySimplex, 1, new Scalar(0, 0, 255), 2);

}

pictureBox2.Image = new Bitmap(result_image.ToMemoryStream());

textBox1.Text = sb.ToString();

}

else

{

textBox1.Text = "未检测到人脸";

}

}

private void pictureBox2_DoubleClick(object sender, EventArgs e)

{

Common.ShowNormalImg(pictureBox2.Image);

}

private void pictureBox1_DoubleClick(object sender, EventArgs e)

{

Common.ShowNormalImg(pictureBox1.Image);

}

}

}

FreeYoloFace.cs

using OpenCvSharp.Dnn;
using OpenCvSharp;
using System;
using System.Collections.Generic;
using System.Linq;

namespace OpenCvSharp_DNN_Demo
{
    public class FreeYoloFace
    {

        float confThreshold;
        float nmsThreshold;

        int num_stride = 3;
        float[] strides = new float[3] { 8.0f, 16.0f, 32.0f };

        string modelpath;

        int inpHeight;
        int inpWidth;

        List<string> class_names;
        int num_class;

        Net opencv_net;
        Mat BN_image;

        Mat image;

        public FreeYoloFace(string modelpath)
        {
            opencv_net = CvDnn.ReadNetFromOnnx(modelpath);

            class_names = new List<string> { "face" };
            num_class = 1;

            confThreshold = 0.8f;
            nmsThreshold = 0.5f;

            inpHeight = 192;
            inpWidth = 320;
        }


        unsafe public List<Face> Detect(Mat image)
        {
            List<Face> ltFace = new List<Face>();
            float ratio = Math.Min(1.0f * inpHeight / image.Rows, 1.0f * inpWidth / image.Cols);
            int neww = (int)(image.Cols * ratio);
            int newh = (int)(image.Rows * ratio);

            Mat dstimg = new Mat();
            Cv2.Resize(image, dstimg, new OpenCvSharp.Size(neww, newh));
            Cv2.CopyMakeBorder(dstimg, dstimg, 0, inpHeight - newh, 0, inpWidth - neww, BorderTypes.Constant);
            BN_image = CvDnn.BlobFromImage(dstimg);

            //配置图片输入数据
            opencv_net.SetInput(BN_image);

            //模型推理,读取推理结果
            Mat[] outs = new Mat[1] { new Mat() };
            string[] outBlobNames = opencv_net.GetUnconnectedOutLayersNames().ToArray();

            opencv_net.Forward(outs, outBlobNames);

            int num_proposal = outs[0].Size(1);
            int nout = outs[0].Size(2);

            float* pdata = (float*)outs[0].Data;

            List<float> confidences = new List<float>();
            List<Rect> boxes = new List<Rect>();
            List<int> classIds = new List<int>();

            for (int n = 0; n < num_stride; n++)
            {
                int num_grid_x = (int)Math.Ceiling(inpWidth / strides[n]);
                int num_grid_y = (int)Math.Ceiling(inpHeight / strides[n]);

                for (int i = 0; i < num_grid_y; i++)
                {
                    for (int j = 0; j < num_grid_x; j++)
                    {
                        float box_score = pdata[4];
                        int max_ind = 0;
                        float max_class_socre = 0;
                        for (int k = 0; k < num_class; k++)
                        {
                            if (pdata[k + 5] > max_class_socre)
                            {
                                max_class_socre = pdata[k + 5];
                                max_ind = k;
                            }
                        }
                        max_class_socre = max_class_socre * box_score;
                        max_class_socre = (float)Math.Sqrt(max_class_socre);

                        if (max_class_socre > confThreshold)
                        {
                            float cx = (0.5f + j + pdata[0]) * strides[n];  //cx
                            float cy = (0.5f + i + pdata[1]) * strides[n];   //cy
                            float w = (float)(Math.Exp(pdata[2]) * strides[n]);   //w
                            float h = (float)(Math.Exp(pdata[3]) * strides[n]);  //h

                            float xmin = (float)((cx - 0.5 * w) / ratio);
                            float ymin = (float)((cy - 0.5 * h) / ratio);
                            float xmax = (float)((cx + 0.5 * w) / ratio);
                            float ymax = (float)((cy + 0.5 * h) / ratio);

                            int left = (int)((cx - 0.5 * w) / ratio);
                            int top = (int)((cy - 0.5 * h) / ratio);
                            int width = (int)(w / ratio);
                            int height = (int)(h / ratio);

                            confidences.Add(max_class_socre);
                            boxes.Add(new Rect(left, top, width, height));
                            classIds.Add(max_ind);
                        }
                        pdata += nout;
                    }
                }

            }

            int[] indices;
            CvDnn.NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, out indices);

            for (int ii = 0; ii < indices.Length; ++ii)
            {
                int idx = indices[ii];
                Rect box = boxes[idx];
                ltFace.Add(new Face(box, confidences[idx]));
            }

            outs[0].Dispose();
            BN_image.Dispose();
            dstimg.Dispose();

            return ltFace;
        }
    }
}

FaceQualityAssessment.cs

using OpenCvSharp;
using OpenCvSharp.Dnn;
using System.Linq;

namespace OpenCvSharp_DNN_Demo
{
    public class FaceQualityAssessment
    {
        Net net;

        int inpWidth = 112;
        int inpHeight = 112;

        float[] mean = new float[] { 0.5f, 0.5f, 0.5f };

        float[] std = new float[] { 0.5f, 0.5f, 0.5f };

        public FaceQualityAssessment(string modelpath)
        {
            net = CvDnn.ReadNetFromOnnx(modelpath);
        }

        unsafe public float Detect(Mat cropped)
        {
            Mat rgbimg = new Mat();
            Cv2.CvtColor(cropped, rgbimg, ColorConversionCodes.BGR2RGB);
            Cv2.Resize(rgbimg, rgbimg, new Size(inpWidth, inpHeight));
            Mat normalized_mat = Normalize(rgbimg);

            Mat blob = CvDnn.BlobFromImage(normalized_mat);

            //配置图片输入数据
            net.SetInput(blob);

            //模型推理,读取推理结果
            Mat[] outs = new Mat[1] { new Mat() };
            string[] outBlobNames = net.GetUnconnectedOutLayersNames().ToArray();

            net.Forward(outs, outBlobNames);

            float* pdata = (float*)outs[0].Data;  //形状1x10
            int length = outs[0].Size(1);
            float fqa_prob_mean = 0;
            for (int i = 0; i < length; i++)
            {
                fqa_prob_mean += pdata[i];
            }
            fqa_prob_mean /= length;

            rgbimg.Dispose();
            normalized_mat.Dispose();
            blob.Dispose();
            outs[0].Dispose();
            return fqa_prob_mean;
        }

        Mat Normalize(Mat src)
        {
            Mat[] bgr = src.Split();
            for (int i = 0; i < bgr.Length; ++i)
            {
                bgr[i].ConvertTo(bgr[i], MatType.CV_32FC1, 1.0 / (255.0 * std[i]), (0.0 - mean[i]) / std[i]);
            }
            Cv2.Merge(bgr, src);
            foreach (Mat channel in bgr)
            {
                channel.Dispose();
            }
            return src;
        }

    }
}

下载

源码下载

相关推荐
果冻人工智能32 分钟前
2025 年将颠覆商业的 8 大 AI 应用场景
人工智能·ai员工
代码不行的搬运工33 分钟前
神经网络12-Time-Series Transformer (TST)模型
人工智能·神经网络·transformer
石小石Orz35 分钟前
Three.js + AI:AI 算法生成 3D 萤火虫飞舞效果~
javascript·人工智能·算法
孤独且没人爱的纸鹤1 小时前
【深度学习】:从人工神经网络的基础原理到循环神经网络的先进技术,跨越智能算法的关键发展阶段及其未来趋势,探索技术进步与应用挑战
人工智能·python·深度学习·机器学习·ai
阿_旭1 小时前
TensorFlow构建CNN卷积神经网络模型的基本步骤:数据处理、模型构建、模型训练
人工智能·深度学习·cnn·tensorflow
羊小猪~~1 小时前
tensorflow案例7--数据增强与测试集, 训练集, 验证集的构建
人工智能·python·深度学习·机器学习·cnn·tensorflow·neo4j
极客代码1 小时前
【Python TensorFlow】进阶指南(续篇三)
开发语言·人工智能·python·深度学习·tensorflow
zhangfeng11331 小时前
pytorch 的交叉熵函数,多分类,二分类
人工智能·pytorch·分类
Seeklike1 小时前
11.22 深度学习-pytorch自动微分
人工智能·pytorch·深度学习
庞传奇1 小时前
TensorFlow 的基本概念和使用场景
人工智能·python·tensorflow