效果
![0a423fa3504d8512cdf67535afc9fcc0.png](https://img-blog.csdnimg.cn/img_convert/0a423fa3504d8512cdf67535afc9fcc0.png)
模型信息
Inputs
-------------------------
name:input
tensor:Float[1, 3, 256, 384]
---------------------------------------------------------------
Outputs
-------------------------
name:regression
tensor:Float[1, 18414, 4]
name:classification
tensor:Float[1, 18414, 1]
name:segmentation
tensor:Float[1, 3, 256, 384]
---------------------------------------------------------------
项目
![41231063f6631822dd31c70ab23f4f10.png](https://img-blog.csdnimg.cn/img_convert/41231063f6631822dd31c70ab23f4f10.png)
代码
using OpenCvSharp;
using OpenCvSharp.Dnn;
using System;
using System.Collections.Generic;
using System.Drawing;
using System.IO;
using System.Linq;
using System.Numerics;
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;
float confThreshold;
float nmsThreshold;
string modelpath;
string anchorpath;
int inpHeight;
int inpWidth;
float[] mean = { 0.485f, 0.456f, 0.406f };
float[] std = { 0.229f, 0.224f, 0.225f };
List<string> det_class_names = new List<string>() { "car" };
List<string> seg_class_names = new List<string>() { "Background", "Lane", "Line" };
List<Vec3b> class_colors = new List<Vec3b> { new Vec3b(0, 0, 0), new Vec3b(0, 255, 0), new Vec3b(255, 0, 0) };
int det_num_class = 1;
int seg_numclass = 3;
float[] anchors;
Net opencv_net;
Mat BN_image;
Mat image;
Mat result_image;
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)
{
confThreshold = 0.3f;
nmsThreshold = 0.5f;
modelpath = "model/hybridnets_256x384.onnx";
anchorpath = "model/anchors_73656.bin";
inpHeight = 256;
inpWidth = 384;
opencv_net = CvDnn.ReadNetFromOnnx(modelpath);
FileStream fileStream = new FileStream(anchorpath, FileMode.Open);
//读二进制文件类
BinaryReader br = new BinaryReader(fileStream, Encoding.UTF8);
int len = 73656;
anchors = new float[len];
byte[] byteTemp;
float fTemp;
for (int i = 0; i < len; i++)
{
byteTemp = br.ReadBytes(4);
fTemp = BitConverter.ToSingle(byteTemp, 0);
anchors[i] = fTemp;
}
br.Close();
image_path = "test_img/test.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……";
pictureBox2.Image = null;
Application.DoEvents();
image = new Mat(image_path);
int newh = 0, neww = 0, padh = 0, padw = 0;
Mat resize_img = Common.ResizeImage(image, inpHeight, inpWidth, ref newh, ref neww, ref padh, ref padw);
float ratioh = (float)image.Rows / newh;
float ratiow = (float)image.Cols / neww;
Mat normalize = Common.Normalize(resize_img, mean, std);
dt1 = DateTime.Now;
BN_image = CvDnn.BlobFromImage(normalize);
//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[3] { new Mat(), new Mat(), new Mat() };
string[] outBlobNames = opencv_net.GetUnconnectedOutLayersNames().ToArray();
opencv_net.Forward(outs, outBlobNames);
dt2 = DateTime.Now;
float* classification = (float*)outs[0].Data;
float* box_regression = (float*)outs[1].Data;
float* seg = (float*)outs[2].Data;
List<Rect> boxes = new List<Rect>();
List<float> confidences = new List<float>();
List<int> classIds = new List<int>();
int num_proposal = outs[1].Size(1); //输入的是单张图, 第0维batchsize忽略
for (int n = 0; n < num_proposal; n++)
{
float conf = classification[n];
if (conf > confThreshold)
{
int row_ind = n * 4;
float x_centers = box_regression[row_ind + 1] * anchors[row_ind + 2] + anchors[row_ind];
float y_centers = box_regression[row_ind] * anchors[row_ind + 3] + anchors[row_ind + 1];
float w = (float)(Math.Exp(box_regression[row_ind + 3]) * anchors[row_ind + 2]);
float h = (float)(Math.Exp(box_regression[row_ind + 2]) * anchors[row_ind + 3]);
float xmin = (float)((x_centers - w * 0.5 - padw) * ratiow);
float ymin = (float)((y_centers - h * 0.5 - padh) * ratioh);
w *= ratiow;
h *= ratioh;
Rect box = new Rect((int)xmin, (int)ymin, (int)w, (int)h);
boxes.Add(box);
confidences.Add(conf);
classIds.Add(0);
}
}
int[] indices;
CvDnn.NMSBoxes(boxes, confidences, confThreshold, nmsThreshold, out indices);
result_image = image.Clone();
for (int ii = 0; ii < indices.Length; ++ii)
{
int idx = indices[ii];
Rect box = boxes[idx];
Cv2.Rectangle(result_image, new OpenCvSharp.Point(box.X, box.Y), new OpenCvSharp.Point(box.X + box.Width, box.Y + box.Height), new Scalar(0, 0, 255), 2);
string label = det_class_names[classIds[idx]] + ":" + confidences[idx].ToString("0.00");
Cv2.PutText(result_image, label, new OpenCvSharp.Point(box.X, box.Y - 5), HersheyFonts.HersheySimplex, 0.75, new Scalar(0, 0, 255), 1);
}
int area = inpHeight * inpWidth;
int i = 0, j = 0, c = 0;
for (i = 0; i < result_image.Rows; i++)
{
for (j = 0; j < result_image.Cols; j++)
{
int x = (int)((j / ratiow) + padw); ///从原图映射回到输出特征图
int y = (int)((i / ratioh) + padh);
int max_id = -1;
float max_conf = -10000;
for (c = 0; c < seg_numclass; c++)
{
float seg_conf = seg[c * area + y * inpWidth + x];
if (seg_conf > max_conf)
{
max_id = c;
max_conf = seg_conf;
}
}
if (max_id > 0)
{
result_image.Set<Vec3b>(i, j, class_colors[max_id]);
}
}
}
pictureBox2.Image = new Bitmap(result_image.ToMemoryStream());
textBox1.Text = "推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms";
}
private void pictureBox2_DoubleClick(object sender, EventArgs e)
{
Common.ShowNormalImg(pictureBox2.Image);
}
private void pictureBox1_DoubleClick(object sender, EventArgs e)
{
Common.ShowNormalImg(pictureBox1.Image);
}
}
}