devbox
Gausst松鼠会
这个屌丝很懒，什么也没留下！
关注作者
热门标签
热门文章
当前位置:   article > 正文

C# PaddleInference OCR文字识别（只识别）_c# ocr识别

作者：Gausst松鼠会 | 2024-06-09 05:25:41

赞

踩

c# ocr识别

目录

说明

效果

项目

测试图片

代码

下载 

说明

C# PaddleInference OCR文字识别（只识别），没有文字区域检测、文字方向判断

测试图片是文字区域检测裁剪出来、处理过的图片

完整的OCR识别查看 C# PaddleInference OCR识别 学习研究Demo_天天代码码天天的博客-CSDN博客

效果

项目

VS2022+.net4.8+OpenCvSharp4+Sdcb.PaddleInference

测试图片

代码

using OpenCvSharp;
using Sdcb.PaddleInference;
using Sdcb.PaddleInference.Native;
using System;
using System.Collections.Generic;
using System.Drawing;
using System.Globalization;
using System.IO;
using System.Runtime.InteropServices;
using System.Text;
using System.Windows.Forms;

namespace PaddleInference_OCR文字识别
{
    public partial class Form1 : Form
    {
        public Form1()
        {
            InitializeComponent();
        }

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

        OcrShape recShape = new OcrShape(3, 320, 48);
        PaddlePredictor rec_predictor;

        public IReadOnlyList<string> Labels;

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


        private void button1_Click(object sender, EventArgs e)
        {
            OpenFileDialog ofd = new OpenFileDialog();
            ofd.Filter = fileFilter;
            if (ofd.ShowDialog() != DialogResult.OK) return;

            pictureBox1.Image = null;

            img = ofd.FileName;
            bmp = new Bitmap(img);
            pictureBox1.Image = new Bitmap(img);
            textBox1.Text = "";
        }

        private unsafe void Form1_Load(object sender, EventArgs e)
        {

            string startupPath = Application.StartupPath;
            IntPtr det_ptr = PaddleNative.PD_ConfigCreate();
            Encoding PaddleEncoding = Environment.OSVersion.Platform == PlatformID.Win32NT ? Encoding.GetEncoding(CultureInfo.CurrentCulture.TextInfo.ANSICodePage) : Encoding.UTF8;

            //识别模型
            IntPtr rec_ptr = PaddleNative.PD_ConfigCreate();
            String rec_programPath = startupPath + "\\ch_PP-OCRv3_rec\\inference.pdmodel";
            String rec_paramsPath = startupPath + "\\ch_PP-OCRv3_rec\\inference.pdiparams";
            byte[] rec_programBytes = PaddleEncoding.GetBytes(rec_programPath);
            byte[] rec_paramsBytes = PaddleEncoding.GetBytes(rec_paramsPath);
            fixed (byte* rec_programPtr = rec_programBytes)
            fixed (byte* rec_paramsPtr = rec_paramsBytes)
            {
                PaddleNative.PD_ConfigSetModel(rec_ptr, (IntPtr)rec_programPtr, (IntPtr)rec_paramsPtr);
            }

            rec_predictor = new PaddlePredictor(PaddleNative.PD_PredictorCreate(rec_ptr));

            //Labels
            String labelsPath = startupPath + "\\ppocr_keys.txt";
            Stream Steam = new FileStream(labelsPath, FileMode.Open, FileAccess.Read, FileShare.Read);
            StreamReader reader = new StreamReader(Steam);
            List<string> tempList = new List<string>();
            while (!reader.EndOfStream)
            {
                tempList.Add(reader.ReadLine());
            }
            reader.Dispose();
            Steam.Dispose();
            Labels = tempList;
        }

        private void button3_Click(object sender, EventArgs e)
        {
            if (pictureBox1.Image == null)
            {
                return;
            }
            dt1 = DateTime.Now;
            var src = OpenCvSharp.Extensions.BitmapConverter.ToMat(bmp);

            int modelHeight = recShape.Height;
            int maxWidth = (int)Math.Ceiling(1.0 * src.Width / src.Height * modelHeight);

            Mat channel3 = new Mat();
            if (src.Channels() == 4)
            {
                channel3 = src.CvtColor(ColorConversionCodes.RGBA2BGR);
            }
            else if (src.Channels() == 3)
            {
                channel3 = src.Clone();
            }
            else if (src.Channels() == 1)
            {
                channel3 = src.CvtColor(ColorConversionCodes.GRAY2RGB);
            }
            else
            {
                throw new Exception("Unexpect src channel: {" + src.Channels() + "}, allow: (1/3/4)");
            }

            Mat resized = ResizePadding(channel3, modelHeight, maxWidth);
            Mat normalize = Normalize(resized);

            using (PaddleTensor input = rec_predictor.GetInputTensor(rec_predictor.InputNames[0]))
            {
                int channel = normalize.Channels();
                input.Shape = new[] { 1, channel, modelHeight, maxWidth };
                float[] data = ExtractMat(normalize, channel, modelHeight, maxWidth);
                input.SetData(data);
            }

            normalize.Dispose();
            resized.Dispose();

            if (!rec_predictor.Run())
            {
                throw new Exception($"PaddlePredictor(Recognizer) run failed.");
            }

            using (PaddleTensor output = rec_predictor.GetOutputTensor(rec_predictor.OutputNames[0]))
            {
                float[] data = output.GetData<float>();
                int[] shape = output.Shape;
                GCHandle dataHandle = default;
                try
                {
                    dataHandle = GCHandle.Alloc(data, GCHandleType.Pinned);
                    IntPtr dataPtr = dataHandle.AddrOfPinnedObject();

                    int labelCount = shape[2];
                    int charCount = shape[1];

                    StringBuilder sbInfo = new StringBuilder();

                    for (int i = 0; i < shape[0]; i++)
                    {
                        StringBuilder sb = new StringBuilder();
                        int lastIndex = 0;
                        float score = 0;
                        for (int n = 0; n < charCount; ++n)
                        {
                            Mat mat = new Mat(1, labelCount, MatType.CV_32FC1, dataPtr + (n + i * charCount) * labelCount * sizeof(float));
                            int[] maxIdx = new int[2];
                            mat.MinMaxIdx(out double _, out double maxVal, new int[0], maxIdx);
                            if (maxIdx[1] > 0 && (!(n > 0 && maxIdx[1] == lastIndex)))
                            {
                                score += (float)maxVal;
                                sb.Append(GetLabelByIndex(maxIdx[1]));
                            }
                            lastIndex = maxIdx[1];
                            mat.Dispose();
                        }
                        sbInfo.AppendLine("Text:" + sb.ToString());
                        sbInfo.AppendLine("Score:" + score / sb.Length);
                    }

                    dt2 = DateTime.Now;
                    sbInfo.AppendLine("-----------------------------------\n");
                    sbInfo.AppendLine(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss.fff"));
                    sbInfo.AppendLine("耗时:" + (dt2 - dt1).TotalMilliseconds + "ms\n");
                    textBox1.Text = sbInfo.ToString();
                }
                finally
                {
                    dataHandle.Free();
                }
            }

        }

        string GetLabelByIndex(int x)
        {
            if (x > 0 && x <= Labels.Count)
            {
                return Labels[x - 1];
            }
            else if (x == Labels.Count + 1)
            {
                return "";
            }
            else
            {
                throw new Exception("Unable to GetLabelByIndex: index {" + x + "} out of range {" + Labels.Count + "}, OCR model or labels not matched?");
            }
        }

        private Mat ResizePadding(Mat src, int height, int targetWidth)
        {
            OpenCvSharp.Size size = src.Size();
            float whRatio = 1.0f * size.Width / size.Height;
            int width = (int)Math.Ceiling(height * whRatio);

            if (width == targetWidth)
            {
                return src.Resize(new OpenCvSharp.Size(width, height));
            }
            else
            {
                Mat resized = src.Resize(new OpenCvSharp.Size(width, height));
                return resized.CopyMakeBorder(0, 0, 0, targetWidth - width, BorderTypes.Constant, Scalar.Gray);
            }
        }

        private Mat Normalize(Mat src)
        {
            Mat normalized = new Mat();
            src.ConvertTo(normalized, MatType.CV_32FC3, 1.0 / 255);
            Mat[] bgr = normalized.Split();
            float[] scales = new[] { 1 / 0.229f, 1 / 0.224f, 1 / 0.225f };
            float[] means = new[] { 0.485f, 0.456f, 0.406f };
            for (int i = 0; i < bgr.Length; ++i)
            {
                bgr[i].ConvertTo(bgr[i], MatType.CV_32FC1, 1.0 * scales[i], (0.0 - means[i]) * scales[i]);
            }
            normalized.Dispose();
            Mat dest = new Mat();
            Cv2.Merge(bgr, dest);
            foreach (Mat channel in bgr)
            {
                channel.Dispose();
            }
            return dest;
        }

        private float[] ExtractMat(Mat mat, int channel, int height, int width)
        {
            float[] result = new float[1 * channel * width * height];
            GCHandle resultHandle = GCHandle.Alloc(result, GCHandleType.Pinned);
            IntPtr resultPtr = resultHandle.AddrOfPinnedObject();
            try
            {

                Mat src = mat.Clone();
                if (src.Channels() != channel)
                {
                    throw new Exception($"src channel={src.Channels()}, expected {channel}");
                }
                for (int c = 0; c < channel; ++c)
                {
                    Mat dest = new Mat(height, width, MatType.CV_32FC1, resultPtr + c * height * width * sizeof(float));
                    Cv2.ExtractChannel(src, dest, c);
                    dest.Dispose();
                }
                return result;
            }
            finally
            {
                resultHandle.Free();
            }
        }

        private float[] ExtractMat(Mat[] srcs, int channel, int height, int width)
        {
            float[] result = new float[srcs.Length * channel * width * height];
            GCHandle resultHandle = GCHandle.Alloc(result, GCHandleType.Pinned);
            IntPtr resultPtr = resultHandle.AddrOfPinnedObject();
            try
            {
                for (int i = 0; i < srcs.Length; ++i)
                {
                    Mat src = srcs[i];
                    if (src.Channels() != channel)
                    {
                        throw new Exception($"src[{i}] channel={src.Channels()}, expected {channel}");
                    }
                    for (int c = 0; c < channel; ++c)
                    {
                        Mat dest = new Mat(height, width, MatType.CV_32FC1, resultPtr + (c + i * channel) * height * width * sizeof(float));
                        Cv2.ExtractChannel(src, dest, c);
                        dest.Dispose();
                    }
                }
                return result;
            }
            finally
            {
                resultHandle.Free();
            }
        }

    }
}
 

  1. using OpenCvSharp;
  2. using Sdcb.PaddleInference;
  3. using Sdcb.PaddleInference.Native;
  4. using System;
  5. using System.Collections.Generic;
  6. using System.Drawing;
  7. using System.Globalization;
  8. using System.IO;
  9. using System.Runtime.InteropServices;
  10. using System.Text;
  11. using System.Windows.Forms;
  12.  
  13. namespace PaddleInference_OCR文字识别
  14. {
  15.     public partial class Form1 : Form
  16.     {
  17.         public Form1()
  18.         {
  19.             InitializeComponent();
  20.         }
  21.  
  22.         Bitmap bmp;
  23.         string fileFilter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";
  24.         string img = "";
  25.         string startupPath = "";
  26.  
  27.         OcrShape recShape = new OcrShape(3, 320, 48);
  28.         PaddlePredictor rec_predictor;
  29.  
  30.         public IReadOnlyList<string> Labels;
  31.  
  32.         DateTime dt1 = DateTime.Now;
  33.         DateTime dt2 = DateTime.Now;
  34.  
  35.  
  36.         private void button1_Click(object sender, EventArgs e)
  37.         {
  38.             OpenFileDialog ofd = new OpenFileDialog();
  39.             ofd.Filter = fileFilter;
  40.             if (ofd.ShowDialog() != DialogResult.OK) return;
  41.  
  42.             pictureBox1.Image = null;
  43.  
  44.             img = ofd.FileName;
  45.             bmp = new Bitmap(img);
  46.             pictureBox1.Image = new Bitmap(img);
  47.             textBox1.Text = "";
  48.         }
  49.  
  50.  
  51.  
  52.         private unsafe void Form1_Load(object sender, EventArgs e)
  53.         {
  54.  
  55.             string startupPath = Application.StartupPath;
  56.             IntPtr det_ptr = PaddleNative.PD_ConfigCreate();
  57.             Encoding PaddleEncoding = Environment.OSVersion.Platform == PlatformID.Win32NT ? Encoding.GetEncoding(CultureInfo.CurrentCulture.TextInfo.ANSICodePage) : Encoding.UTF8;
  58.  
  59.             //识别模型
  60.             IntPtr rec_ptr = PaddleNative.PD_ConfigCreate();
  61.             String rec_programPath = startupPath + "\\ch_PP-OCRv3_rec\\inference.pdmodel";
  62.             String rec_paramsPath = startupPath + "\\ch_PP-OCRv3_rec\\inference.pdiparams";
  63.             byte[] rec_programBytes = PaddleEncoding.GetBytes(rec_programPath);
  64.             byte[] rec_paramsBytes = PaddleEncoding.GetBytes(rec_paramsPath);
  65.             fixed (byte* rec_programPtr = rec_programBytes)
  66.             fixed (byte* rec_paramsPtr = rec_paramsBytes)
  67.             {
  68.                 PaddleNative.PD_ConfigSetModel(rec_ptr, (IntPtr)rec_programPtr, (IntPtr)rec_paramsPtr);
  69.             }
  70.  
  71.             rec_predictor = new PaddlePredictor(PaddleNative.PD_PredictorCreate(rec_ptr));
  72.  
  73.             //Labels
  74.             String labelsPath = startupPath + "\\ppocr_keys.txt";
  75.             Stream Steam = new FileStream(labelsPath, FileMode.Open, FileAccess.Read, FileShare.Read);
  76.             StreamReader reader = new StreamReader(Steam);
  77.             List<string> tempList = new List<string>();
  78.             while (!reader.EndOfStream)
  79.             {
  80.                 tempList.Add(reader.ReadLine());
  81.             }
  82.             reader.Dispose();
  83.             Steam.Dispose();
  84.             Labels = tempList;
  85.         }
  86.  
  87.         private void button3_Click(object sender, EventArgs e)
  88.         {
  89.             if (pictureBox1.Image == null)
  90.             {
  91.                 return;
  92.             }
  93.             dt1 = DateTime.Now;
  94.             var src = OpenCvSharp.Extensions.BitmapConverter.ToMat(bmp);
  95.  
  96.             int modelHeight = recShape.Height;
  97.             int maxWidth = (int)Math.Ceiling(1.0 * src.Width / src.Height * modelHeight);
  98.  
  99.             Mat channel3 = new Mat();
  100.             if (src.Channels() == 4)
  101.             {
  102.                 channel3 = src.CvtColor(ColorConversionCodes.RGBA2BGR);
  103.             }
  104.             else if (src.Channels() == 3)
  105.             {
  106.                 channel3 = src.Clone();
  107.             }
  108.             else if (src.Channels() == 1)
  109.             {
  110.                 channel3 = src.CvtColor(ColorConversionCodes.GRAY2RGB);
  111.             }
  112.             else
  113.             {
  114.                 throw new Exception("Unexpect src channel: {" + src.Channels() + "}, allow: (1/3/4)");
  115.             }
  116.  
  117.             Mat resized = ResizePadding(channel3, modelHeight, maxWidth);
  118.             Mat normalize = Normalize(resized);
  119.  
  120.             using (PaddleTensor input = rec_predictor.GetInputTensor(rec_predictor.InputNames[0]))
  121.             {
  122.                 int channel = normalize.Channels();
  123.                 input.Shape = new[] { 1, channel, modelHeight, maxWidth };
  124.                 float[] data = ExtractMat(normalize, channel, modelHeight, maxWidth);
  125.                 input.SetData(data);
  126.             }
  127.  
  128.             normalize.Dispose();
  129.             resized.Dispose();
  130.  
  131.             if (!rec_predictor.Run())
  132.             {
  133.                 throw new Exception($"PaddlePredictor(Recognizer) run failed.");
  134.             }
  135.  
  136.             using (PaddleTensor output = rec_predictor.GetOutputTensor(rec_predictor.OutputNames[0]))
  137.             {
  138.                 float[] data = output.GetData<float>();
  139.                 int[] shape = output.Shape;
  140.                 GCHandle dataHandle = default;
  141.                 try
  142.                 {
  143.                     dataHandle = GCHandle.Alloc(data, GCHandleType.Pinned);
  144.                     IntPtr dataPtr = dataHandle.AddrOfPinnedObject();
  145.  
  146.                     int labelCount = shape[2];
  147.                     int charCount = shape[1];
  148.  
  149.                     StringBuilder sbInfo = new StringBuilder();
  150.  
  151.                     for (int i = 0; i < shape[0]; i++)
  152.                     {
  153.                         StringBuilder sb = new StringBuilder();
  154.                         int lastIndex = 0;
  155.                         float score = 0;
  156.                         for (int n = 0; n < charCount; ++n)
  157.                         {
  158.                             Mat mat = new Mat(1, labelCount, MatType.CV_32FC1, dataPtr + (n + i * charCount) * labelCount * sizeof(float));
  159.                             int[] maxIdx = new int[2];
  160.                             mat.MinMaxIdx(out double _, out double maxVal, new int[0], maxIdx);
  161.                             if (maxIdx[1] > 0 && (!(n > 0 && maxIdx[1] == lastIndex)))
  162.                             {
  163.                                 score += (float)maxVal;
  164.                                 sb.Append(GetLabelByIndex(maxIdx[1]));
  165.                             }
  166.                             lastIndex = maxIdx[1];
  167.                             mat.Dispose();
  168.                         }
  169.                         sbInfo.AppendLine("Text:" + sb.ToString());
  170.                         sbInfo.AppendLine("Score:" + score / sb.Length);
  171.                     }
  172.  
  173.                     dt2 = DateTime.Now;
  174.                     sbInfo.AppendLine("-----------------------------------\n");
  175.                     sbInfo.AppendLine(DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss.fff"));
  176.                     sbInfo.AppendLine("耗时:" + (dt2 - dt1).TotalMilliseconds + "ms\n");
  177.                     textBox1.Text = sbInfo.ToString();
  178.                 }
  179.                 finally
  180.                 {
  181.                     dataHandle.Free();
  182.                 }
  183.             }
  184.  
  185.         }
  186.  
  187.         string GetLabelByIndex(int x)
  188.         {
  189.             if (x > 0 && x <= Labels.Count)
  190.             {
  191.                 return Labels[x - 1];
  192.             }
  193.             else if (x == Labels.Count + 1)
  194.             {
  195.                 return "";
  196.             }
  197.             else
  198.             {
  199.                 throw new Exception("Unable to GetLabelByIndex: index {" + x + "} out of range {" + Labels.Count + "}, OCR model or labels not matched?");
  200.             }
  201.         }
  202.  
  203.         private Mat ResizePadding(Mat src, int height, int targetWidth)
  204.         {
  205.             OpenCvSharp.Size size = src.Size();
  206.             float whRatio = 1.0f * size.Width / size.Height;
  207.             int width = (int)Math.Ceiling(height * whRatio);
  208.  
  209.             if (width == targetWidth)
  210.             {
  211.                 return src.Resize(new OpenCvSharp.Size(width, height));
  212.             }
  213.             else
  214.             {
  215.                 Mat resized = src.Resize(new OpenCvSharp.Size(width, height));
  216.                 return resized.CopyMakeBorder(0, 0, 0, targetWidth - width, BorderTypes.Constant, Scalar.Gray);
  217.             }
  218.         }
  219.  
  220.         private Mat Normalize(Mat src)
  221.         {
  222.             Mat normalized = new Mat();
  223.             src.ConvertTo(normalized, MatType.CV_32FC3, 1.0 / 255);
  224.             Mat[] bgr = normalized.Split();
  225.             float[] scales = new[] { 1 / 0.229f, 1 / 0.224f, 1 / 0.225f };
  226.             float[] means = new[] { 0.485f, 0.456f, 0.406f };
  227.             for (int i = 0; i < bgr.Length; ++i)
  228.             {
  229.                 bgr[i].ConvertTo(bgr[i], MatType.CV_32FC1, 1.0 * scales[i], (0.0 - means[i]) * scales[i]);
  230.             }
  231.             normalized.Dispose();
  232.             Mat dest = new Mat();
  233.             Cv2.Merge(bgr, dest);
  234.             foreach (Mat channel in bgr)
  235.             {
  236.                 channel.Dispose();
  237.             }
  238.             return dest;
  239.         }
  240.  
  241.  
  242.  
  243.         private float[] ExtractMat(Mat mat, int channel, int height, int width)
  244.         {
  245.             float[] result = new float[1 * channel * width * height];
  246.             GCHandle resultHandle = GCHandle.Alloc(result, GCHandleType.Pinned);
  247.             IntPtr resultPtr = resultHandle.AddrOfPinnedObject();
  248.             try
  249.             {
  250.  
  251.                 Mat src = mat.Clone();
  252.                 if (src.Channels() != channel)
  253.                 {
  254.                     throw new Exception($"src channel={src.Channels()}, expected {channel}");
  255.                 }
  256.                 for (int c = 0; c < channel; ++c)
  257.                 {
  258.                     Mat dest = new Mat(height, width, MatType.CV_32FC1, resultPtr + c * height * width * sizeof(float));
  259.                     Cv2.ExtractChannel(src, dest, c);
  260.                     dest.Dispose();
  261.                 }
  262.                 return result;
  263.             }
  264.             finally
  265.             {
  266.                 resultHandle.Free();
  267.             }
  268.         }
  269.  
  270.  
  271.  
  272.         private float[] ExtractMat(Mat[] srcs, int channel, int height, int width)
  273.         {
  274.             float[] result = new float[srcs.Length * channel * width * height];
  275.             GCHandle resultHandle = GCHandle.Alloc(result, GCHandleType.Pinned);
  276.             IntPtr resultPtr = resultHandle.AddrOfPinnedObject();
  277.             try
  278.             {
  279.                 for (int i = 0; i < srcs.Length; ++i)
  280.                 {
  281.                     Mat src = srcs[i];
  282.                     if (src.Channels() != channel)
  283.                     {
  284.                         throw new Exception($"src[{i}] channel={src.Channels()}, expected {channel}");
  285.                     }
  286.                     for (int c = 0; c < channel; ++c)
  287.                     {
  288.                         Mat dest = new Mat(height, width, MatType.CV_32FC1, resultPtr + (c + i * channel) * height * width * sizeof(float));
  289.                         Cv2.ExtractChannel(src, dest, c);
  290.                         dest.Dispose();
  291.                     }
  292.                 }
  293.                 return result;
  294.             }
  295.             finally
  296.             {
  297.                 resultHandle.Free();
  298.             }
  299.         }
  300.  
  301.     }
  302. }

下载 

Demo下载

声明：本文内容由网友自发贡献，不代表【wpsshop博客】立场，版权归原作者所有，本站不承担相应法律责任。如您发现有侵权的内容，请联系我们。转载请注明出处：https://www.wpsshop.cn/w/Gausst松鼠会/article/detail/692794
推荐阅读
相关标签

Copyright © 2003-2013 www.wpsshop.cn 版权所有，并保留所有权利。

  

闽ICP备14008679号