C# OpenCvSharp DNN 实现百度网盘AI大赛-表格检测第2名方案第一部分-表格边界框检测

目录

说明

效果

模型

项目

代码

frmMain.cs

YoloDet.cs 

参考

下载

其他

说明

百度网盘AI大赛-表格检测的第2名方案。

该算法包含表格边界框检测、表格分割和表格方向识别三个部分，首先，ppyoloe-plus-x 对边界框进行预测，并对置信度较高的表格边界框（box）进行裁剪。裁剪后的单个表格实例会送入到DBNet中进行语义分割，分割结果通过opencv轮廓处理获得表格关键点(point)。之后，我们根据DBNet计算的关键点在裁剪后的单个表格实例上绘制表格边界。最后，PP-LCNet结合表格边界先验和表格实例图像，对表格的方向进行预测，并根据之前定义的几何轮廓点与语义轮廓点的对应关系，将几何轮廓点映射为语义轮廓点。

本文使用C# OpenCvSharp DNN 实现百度网盘AI大赛-表格检测第2名方案第一部分-表格边界框检测。

效果

模型

Model Properties
-------------------------
date：2024-10-28T13:52:42.181333
description：Ultralytics YOLO11l model trained on coco.yaml
author：Ultralytics
version：8.3.23
task：detect
license：AGPL-3.0 License (https://ultralytics.com/license)
docs：https://docs.ultralytics.com
stride：32
batch：1
imgsz：[928, 928]
names：{0: 'table'}
---------------------------------------------------------------

Inputs
-------------------------
name：images
tensor：Float[1, 3, 928, 928]
---------------------------------------------------------------

Outputs
-------------------------
name：output0
tensor：Float[1, 5, 17661]
---------------------------------------------------------------

项目

代码

frmMain.cs

using OpenCvSharp;
using System;
using System.Collections.Generic;
using System.Drawing;
using System.Windows.Forms;

namespace OpenCvSharp_DNN_Demo
{
    public partial class frmMain : Form
    {
        public frmMain()
        {
            InitializeComponent();
        }

        YoloDet obj_detector;

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

        DateTime dt1 = DateTime.Now;
        DateTime dt2 = DateTime.Now;

        Mat 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)
        {
            string obj_model_path = "model/yolo_obj_det.onnx";

            obj_detector = new YoloDet(obj_model_path);

            image_path = "test_img/real5.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();

            Mat src = new Mat(image_path);

            dt1 = DateTime.Now;
            List result = obj_detector.infer(src);
            dt2 = DateTime.Now;

            //绘制
            Mat draw_img = src.Clone();
            for (int i = 0; i < result.Count; i++)
            {
                Rect r = Rect.FromLTRB(result[i].xmin, result[i].ymin, result[i].xmax, result[i].ymax);

                Cv2.PutText(draw_img, $"table:{result[0].score:P0}", new OpenCvSharp.Point(r.TopLeft.X, r.TopLeft.Y - 10), HersheyFonts.HersheySimplex, 8, Scalar.Red, 8);
                Cv2.Rectangle(draw_img, r, Scalar.Red, thickness: 8);

                //裁剪，保存，为下一步检测做准备
                Mat crop_img = new Mat(src, r);
                Cv2.ImWrite(i + ".jpg", crop_img);
            }
            pictureBox2.Image = new Bitmap(draw_img.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);
        }
    }
}
 

  1. using OpenCvSharp;
  2. using System;
  3. using System.Collections.Generic;
  4. using System.Drawing;
  5. using System.Windows.Forms;
  6.  
  7. namespace OpenCvSharp_DNN_Demo
  8. {
  9.     public partial class frmMain : Form
  10.     {
  11.         public frmMain()
  12.         {
  13.             InitializeComponent();
  14.         }
  15.  
  16.         YoloDet obj_detector;
  17.  
  18.         string fileFilter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";
  19.         string image_path = "";
  20.  
  21.         DateTime dt1 = DateTime.Now;
  22.         DateTime dt2 = DateTime.Now;
  23.  
  24.         Mat image;
  25.  
  26.         private void button1_Click(object sender, EventArgs e)
  27.         {
  28.             OpenFileDialog ofd = new OpenFileDialog();
  29.             ofd.Filter = fileFilter;
  30.             if (ofd.ShowDialog() != DialogResult.OK) return;
  31.  
  32.             pictureBox1.Image = null;
  33.             pictureBox2.Image = null;
  34.             textBox1.Text = "";
  35.  
  36.             image_path = ofd.FileName;
  37.             pictureBox1.Image = new Bitmap(image_path);
  38.             image = new Mat(image_path);
  39.         }
  40.  
  41.         private void Form1_Load(object sender, EventArgs e)
  42.         {
  43.             string obj_model_path = "model/yolo_obj_det.onnx";
  44.  
  45.             obj_detector = new YoloDet(obj_model_path);
  46.  
  47.             image_path = "test_img/real5.jpg";
  48.             pictureBox1.Image = new Bitmap(image_path);
  49.  
  50.         }
  51.  
  52.         private unsafe void button2_Click(object sender, EventArgs e)
  53.         {
  54.             if (image_path == "")
  55.             {
  56.                 return;
  57.             }
  58.             textBox1.Text = "检测中，请稍等……";
  59.             pictureBox2.Image = null;
  60.             Application.DoEvents();
  61.  
  62.             Mat src = new Mat(image_path);
  63.  
  64.             dt1 = DateTime.Now;
  65.             List<Bbox> result = obj_detector.infer(src);
  66.             dt2 = DateTime.Now;
  67.  
  68.             //绘制
  69.             Mat draw_img = src.Clone();
  70.             for (int i = 0; i < result.Count; i++)
  71.             {
  72.                 Rect r = Rect.FromLTRB(result[i].xmin, result[i].ymin, result[i].xmax, result[i].ymax);
  73.  
  74.                 Cv2.PutText(draw_img, $"table:{result[0].score:P0}", new OpenCvSharp.Point(r.TopLeft.X, r.TopLeft.Y - 10), HersheyFonts.HersheySimplex, 8, Scalar.Red, 8);
  75.                 Cv2.Rectangle(draw_img, r, Scalar.Red, thickness: 8);
  76.  
  77.                 //裁剪，保存，为下一步检测做准备
  78.                 Mat crop_img = new Mat(src, r);
  79.                 Cv2.ImWrite(i + ".jpg", crop_img);
  80.             }
  81.             pictureBox2.Image = new Bitmap(draw_img.ToMemoryStream());
  82.             textBox1.Text = "推理耗时:" + (dt2 - dt1).TotalMilliseconds + "ms";
  83.         }
  84.        
  85.         private void pictureBox2_DoubleClick(object sender, EventArgs e)
  86.         {
  87.             Common.ShowNormalImg(pictureBox2.Image);
  88.         }
  89.         
  90.         private void pictureBox1_DoubleClick(object sender, EventArgs e)
  91.         {
  92.             Common.ShowNormalImg(pictureBox1.Image);
  93.         }
  94.     }
  95. }

YoloDet.cs 

  1. using OpenCvSharp;
  2. using OpenCvSharp.Dnn;
  3. using System;
  4. using System.Collections.Generic;
  5. using System.Linq;
  6.  
  7. namespace OpenCvSharp_DNN_Demo
  8. {
  9.     internal class YoloDet
  10.     {
  11.  
  12.         Net model;
  13.         int[] resize_shape = new int[2] { 928, 928 };
  14.  
  15.         public YoloDet(string model_path)
  16.         {
  17.             model = CvDnn.ReadNetFromOnnx(model_path);
  18.         }
  19.  
  20.         unsafe public List<Bbox> infer(Mat srcimg, float score = 0.4f)
  21.         {
  22.             int ori_h = srcimg.Rows;
  23.             int ori_w = srcimg.Cols;
  24.             img_preprocess
  25.             Mat img;
  26.             int new_w = 0;
  27.             int new_h = 0;
  28.             int left = 0;
  29.             int top = 0;
  30.             img = Common.ResizePad(srcimg, resize_shape[0], ref new_w, ref new_h, ref left, ref top);
  31.  
  32.             //Cv2.ImWrite("0.jpg", img);
  33.  
  34.             img.ConvertTo(img, MatType.CV_32FC3, 1.0 / 255.0);
  35.             Mat blob = CvDnn.BlobFromImage(img);
  36.  
  37.             model.SetInput(blob);
  38.  
  39.             //模型推理，读取推理结果
  40.             Mat[] outs = new Mat[1] { new Mat() };
  41.             string[] outBlobNames = model.GetUnconnectedOutLayersNames().ToArray();
  42.             model.Forward(outs, outBlobNames);
  43.  
  44.             img_postprocess
  45.             float x_factor = (float)ori_w / new_w;
  46.             float y_factor = (float)ori_h / new_h;
  47.             List<Rect> boxes = new List<Rect>();
  48.             List<float> scores = new List<float>();
  49.             int rows = outs[0].Size(2);
  50.  
  51.             //将推理结果转为float数据类型
  52.             //5, 17661
  53.             Mat result_mat = new Mat(5, 17661, MatType.CV_32F, outs[0].Data);
  54.             result_mat = result_mat.T();
  55.             //string str = result_mat.Dump();
  56.  
  57.             for (int i = 0; i < rows; i++)
  58.             {
  59.                 float* ptr = (float*)(result_mat.Ptr(i).ToPointer());
  60.                 float max_score = ptr[4];
  61.                 //float max_score = result_mat.At<float>(i, 4);
  62.  
  63.                 if (max_score >= score)
  64.                 {
  65.                     //At效率不高
  66.                     //float x1 = result_mat.At<float>(i, 0);
  67.                     //float y2 = result_mat.At<float>(i, 1);
  68.                     //float w2 = result_mat.At<float>(i, 2);
  69.                     //float h2 = result_mat.At<float>(i, 3);
  70.  
  71.                     float x = ptr[0];
  72.                     float y = ptr[1];
  73.                     float w = ptr[2];
  74.                     float h = ptr[3];
  75.  
  76.                     int xmin = Math.Max((int)((x - w / 2 - left) * x_factor), 0);
  77.                     int ymin = Math.Max((int)((y - h / 2 - top) * y_factor), 0);
  78.                     boxes.Add(new Rect(xmin, ymin, (int)(w * x_factor), (int)(h * y_factor)));
  79.                     scores.Add(max_score);
  80.                 }
  81.             }
  82.  
  83.             int[] indices;
  84.             CvDnn.NMSBoxes(boxes, scores, score, 0.4f, out indices);
  85.             int num_keep = indices.Length;
  86.             List<Bbox> bboxes = new List<Bbox>();
  87.             for (int i = 0; i < num_keep; i++)
  88.             {
  89.                 int ind = indices[i];
  90.                 bboxes.Add(new Bbox(boxes[ind].X, boxes[ind].Y, Math.Min(boxes[ind].X + boxes[ind].Width, ori_w - 1), Math.Min(boxes[ind].Y + boxes[ind].Height, ori_h - 1), scores[ind]));
  91.             }
  92.             return bboxes;
  93.         }
  94.     }
  95. }
  96.  
  97. /*
  98.  
  99.  Model Properties
  100. -------------------------
  101. date：2024-10-28T13:52:42.181333
  102. description：Ultralytics YOLO11l model trained on coco.yaml
  103. author：Ultralytics
  104. version：8.3.23
  105. task：detect
  106. license：AGPL-3.0 License (https://ultralytics.com/license)
  107. docs：https://docs.ultralytics.com
  108. stride：32
  109. batch：1
  110. imgsz：[928, 928]
  111. names：{0: 'table'}
  112. ---------------------------------------------------------------
  113.  
  114. Inputs
  115. -------------------------
  116. name：images
  117. tensor：Float[1, 3, 928, 928]
  118. ---------------------------------------------------------------
  119.  
  120. Outputs
  121. -------------------------
  122. name：output0
  123. tensor：Float[1, 5, 17661]
  124. ---------------------------------------------------------------
  125.  
  126.  */

参考

  1. https://github.com/hpc203/TableDetection
  2.  
  3. https://aistudio.baidu.com/projectdetail/5398861?searchKeyword=%E8%A1%A8%E6%A0%BC%E6%A3%80%E6%B5%8B%E5%A4%A7%E8%B5%9B&searchTab=ALL

下载

源码下载

其他

C# OnnxRuntime 第二部分-表格分割-CSDN博客

C# OpenCvSharp DNN 第三部分-表格方向识别-CSDN博客

天天代码码天天
微信公众号
.NET 人工智能实践
注：本文转载自blog.csdn.net的天天代码码天天的文章"https://lw112190.blog.csdn.net/article/details/144487387"。版权归原作者所有，此博客不拥有其著作权，亦不承担相应法律责任。如有侵权，请联系我们删除。
复制链接
相关推荐
发表评论

评论记录：

未查询到任何数据！