Sdcb.PaddleOCR 2.7.0.1

The ID prefix of this package has been reserved for one of the owners of this package by NuGet.org. Prefix Reserved
dotnet add package Sdcb.PaddleOCR --version 2.7.0.1
NuGet\Install-Package Sdcb.PaddleOCR -Version 2.7.0.1
This command is intended to be used within the Package Manager Console in Visual Studio, as it uses the NuGet module's version of Install-Package.
<PackageReference Include="Sdcb.PaddleOCR" Version="2.7.0.1" />
For projects that support PackageReference, copy this XML node into the project file to reference the package.
paket add Sdcb.PaddleOCR --version 2.7.0.1
#r "nuget: Sdcb.PaddleOCR, 2.7.0.1"
#r directive can be used in F# Interactive and Polyglot Notebooks. Copy this into the interactive tool or source code of the script to reference the package.
// Install Sdcb.PaddleOCR as a Cake Addin
#addin nuget:?package=Sdcb.PaddleOCR&version=2.7.0.1

// Install Sdcb.PaddleOCR as a Cake Tool
#tool nuget:?package=Sdcb.PaddleOCR&version=2.7.0.1

Sdcb.PaddleOCR

PaddleOCR packages 📖

NuGet Package 💼 Version 📌 Description 📚
Sdcb.PaddleOCR NuGet PaddleOCR library(based on Sdcb.PaddleInference) ⚙️
Sdcb.PaddleOCR.Models.Shared NuGet Shared models/utils for both Online and Local versions 🔗
Sdcb.PaddleOCR.Models.Online NuGet Online PaddleOCR models, will download when first using 🌐
Sdcb.PaddleOCR.Models.Local NuGet Local models, relies on Shared, LocalV3 and LocalV4 models 🏠
Sdcb.PaddleOCR.Models.LocalV3 NuGet Full local v3 models, include multiple language(~105MB) 🗺️
Sdcb.PaddleOCR.Models.LocalV4 NuGet Full local v4 models, include multiple language(~111MB) 🌐

Language supports

Please refer to https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.5/doc/doc_en/models_list_en.md to check language support models.

Just replace the .ChineseV3 in demo code with your speicific language, then you can use the language.

Usage

Windows(Local model): Detection and Recognition(All)

  1. Install NuGet Packages:

    Sdcb.PaddleInference
    Sdcb.PaddleOCR
    Sdcb.PaddleOCR.Models.Local
    Sdcb.PaddleInference.runtime.win64.mkl
    OpenCvSharp4.runtime.win
    
  2. Using following C# code to get result:

    FullOcrModel model = LocalFullModels.ChineseV3;
    
    byte[] sampleImageData;
    string sampleImageUrl = @"https://www.tp-link.com.cn/content/images2017/gallery/4288_1920.jpg";
    using (HttpClient http = new HttpClient())
    {
        Console.WriteLine("Download sample image from: " + sampleImageUrl);
        sampleImageData = await http.GetByteArrayAsync(sampleImageUrl);
    }
    
    using (PaddleOcrAll all = new PaddleOcrAll(model, PaddleDevice.Mkldnn())
    {
        AllowRotateDetection = true, /* 允许识别有角度的文字 */ 
        Enable180Classification = false, /* 允许识别旋转角度大于90度的文字 */
    })
    {
        // Load local file by following code:
        // using (Mat src2 = Cv2.ImRead(@"C:\test.jpg"))
        using (Mat src = Cv2.ImDecode(sampleImageData, ImreadModes.Color))
        {
            PaddleOcrResult result = all.Run(src);
            Console.WriteLine("Detected all texts: \n" + result.Text);
            foreach (PaddleOcrResultRegion region in result.Regions)
            {
                Console.WriteLine($"Text: {region.Text}, Score: {region.Score}, RectCenter: {region.Rect.Center}, RectSize:    {region.Rect.Size}, Angle: {region.Rect.Angle}");
            }
        }
    }
    

Windows(Online model): Detection and Recognition(All)

  1. Install NuGet Packages:

    Sdcb.PaddleInference
    Sdcb.PaddleOCR
    Sdcb.PaddleOCR.Models.Online
    Sdcb.PaddleInference.runtime.win64.mkl
    OpenCvSharp4.runtime.win
    
  2. Using following C# code to get result:

    FullOcrModel model = await OnlineFullModels.EnglishV3.DownloadAsync();
    
    byte[] sampleImageData;
    string sampleImageUrl = @"https://www.tp-link.com.cn/content/images2017/gallery/4288_1920.jpg";
    using (HttpClient http = new HttpClient())
    {
        Console.WriteLine("Download sample image from: " + sampleImageUrl);
        sampleImageData = await http.GetByteArrayAsync(sampleImageUrl);
    }
    
    using (PaddleOcrAll all = new PaddleOcrAll(model, PaddleDevice.Mkldnn())
    {
        AllowRotateDetection = true, /* 允许识别有角度的文字 */ 
        Enable180Classification = false, /* 允许识别旋转角度大于90度的文字 */
    })
    {
        // Load local file by following code:
        // using (Mat src2 = Cv2.ImRead(@"C:\test.jpg"))
        using (Mat src = Cv2.ImDecode(sampleImageData, ImreadModes.Color))
        {
            PaddleOcrResult result = all.Run(src);
            Console.WriteLine("Detected all texts: \n" + result.Text);
            foreach (PaddleOcrResultRegion region in result.Regions)
            {
                Console.WriteLine($"Text: {region.Text}, Score: {region.Score}, RectCenter: {region.Rect.Center}, RectSize:    {region.Rect.Size}, Angle: {region.Rect.Angle}");
            }
        }
    }
    

Linux(Ubuntu 22.04): Detection and Recognition(All)

  1. Use sdflysha/sdflysha/dotnet6-paddle:2.5.0-ubuntu22 to replace mcr.microsoft.com/dotnet/aspnet:6.0 in Dockerfile as docker base image.

The build steps for sdflysha/dotnet6-paddle:2.5.0-ubuntu22 was described here.

  1. Install NuGet Packages:
dotnet add package Sdcb.PaddleOCR.Models.Local

Please aware in Linux, the native binding library is not required, instead, you should compile your own OpenCV/PaddleInference library, or just use the Docker image.

  1. write following C# code to get result(also can be exactly the same as windows):
FullOcrModel model = LocalFullModels.ChineseV3;
using (PaddleOcrAll all = new PaddleOcrAll(model, PaddleDevice.Mkldnn()))
// Load in-memory data by following code:
// using (Mat src = Cv2.ImDecode(sampleImageData, ImreadModes.Color))
using (Mat src = Cv2.ImRead(@"/app/test.jpg"))
{
    Console.WriteLine(all.Run(src).Text);
}

Detection Only

// Install following packages:
// Sdcb.PaddleInference
// Sdcb.PaddleOCR
// Sdcb.PaddleOCR.Models.Local
// Sdcb.PaddleInference.runtime.win64.mkl (required in Windows, linux using docker)
// OpenCvSharp4.runtime.win (required in Windows, linux using docker)
byte[] sampleImageData;
string sampleImageUrl = @"https://www.tp-link.com.cn/content/images2017/gallery/4288_1920.jpg";
using (HttpClient http = new HttpClient())
{
    Console.WriteLine("Download sample image from: " + sampleImageUrl);
    sampleImageData = await http.GetByteArrayAsync(sampleImageUrl);
}

using (PaddleOcrDetector detector = new PaddleOcrDetector(LocalDetectionModel.ChineseV3, PaddleDevice.Mkldnn()))
using (Mat src = Cv2.ImDecode(sampleImageData, ImreadModes.Color))
{
    RotatedRect[] rects = detector.Run(src);
    using (Mat visualized = PaddleOcrDetector.Visualize(src, rects, Scalar.Red, thickness: 2))
    {
        string outputFile = Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.MyPictures), "output.jpg");
        Console.WriteLine("OutputFile: " + outputFile);
        visualized.ImWrite(outputFile);
    }
}

Table recognition

// Install following packages:
// Sdcb.PaddleInference
// Sdcb.PaddleOCR
// Sdcb.PaddleOCR.Models.Local
// Sdcb.PaddleInference.runtime.win64.mkl (required in Windows, linux using docker)
// OpenCvSharp4.runtime.win (required in Windows, linux using docker)
using PaddleOcrTableRecognizer tableRec = new(LocalTableRecognitionModel.ChineseMobileV2_SLANET);
using Mat src = Cv2.ImRead(Path.Combine(Environment.GetFolderPath(Environment.SpecialFolder.MyPictures), "table.jpg"));
// Table detection
TableDetectionResult tableResult = tableRec.Run(src);

// Normal OCR
using PaddleOcrAll all = new(LocalFullModels.ChineseV3);
all.Detector.UnclipRatio = 1.2f;
PaddleOcrResult ocrResult = all.Run(src);

// Rebuild table
string html = tableResult.RebuildTable(ocrResult);
Raw table Table model output Rebuilt table
table image image

Technical details

There is 3 steps to do OCR:

  1. Detection - Detect text's position, angle and area (PaddleOCRDetector)
  2. Classification - Determin whether text should rotate 180 degreee.
  3. Recognization - Recognize the area into text

Optimize parameters and performance hints

PaddleConfig.MkldnnCacheCapacity

Default value: 1

This value has a positive correlation to the peak of memory usage that used by mkldnn and a negative correlation to the performance when providing different images.

To figure out each value corresponding to the peak memory usage, you should run the detection for various images(using the same image will not increase memory usage) continuously till the memory usage get stable within a variation of 1GB.

For more details please check the pr #46 that decreases the default value and the Paddle document for MkldnnCacheCapacity.

PaddleOcrAll.Enable180Classification

Default value: false

This directly effect the step 2, set to false can skip this step, which will unable to detect text from right to left(which should be acceptable because most text direction is from left to right).

Close this option can make the full process about ~10% faster.

PaddleOcrAll.AllowRotateDetection

Default value: true

This allows detect any rotated texts. If your subject is 0 degree text (like scaned table or screenshot), you can set this parameter to false, which will improve OCR accurancy and little bit performance.

PaddleOcrAll.Detector.MaxSize

Default value: 1536

This effect the the max size of step #1, lower this value can improve performance and reduce memory usage, but will also lower the accurancy.

You can also set this value to null, in that case, images will not scale-down to detect, performance will drop and memory will high, but should able to get better accurancy.

How can I improve performance?

Please review the Technical details section and read the Optimize parameters and performance hints section, or UseGpu.

FAQ

How to integrate Sdcb.PaddleOCR to ASP.NET Core?

Please refer to this demo website, it contains a tutorial: https://github.com/sdcb/paddlesharp-ocr-aspnetcore-demo

In your service builder code, register a QueuedPaddleOcrAll Singleton:

builder.Services.AddSingleton(s =>
{
    Action<PaddleConfig> device = builder.Configuration["PaddleDevice"] == "GPU" ? PaddleDevice.Gpu() : PaddleDevice.Mkldnn();
    return new QueuedPaddleOcrAll(() => new PaddleOcrAll(LocalFullModels.ChineseV3, device)
    {
        Enable180Classification = true,
        AllowRotateDetection = true,
    }, consumerCount: 1);
});

In your controller, use the registered QueuedPaddleOcrAll singleton:

public class OcrController : Controller
{
    private readonly QueuedPaddleOcrAll _ocr;

    public OcrController(QueuedPaddleOcrAll ocr) { _ocr = ocr; }

    [Route("ocr")]
    public async Task<OcrResponse> Ocr(IFormFile file)
    {
        using MemoryStream ms = new();
        using Stream stream = file.OpenReadStream();
        stream.CopyTo(ms);
        using Mat src = Cv2.ImDecode(ms.ToArray(), ImreadModes.Color);
        double scale = 1;
        using Mat scaled = src.Resize(default, scale, scale);

        Stopwatch sw = Stopwatch.StartNew();
        string textResult = (await _ocr.Run(scaled)).Text;
        sw.Stop();

        return new OcrResponse(textResult, sw.ElapsedMilliseconds);
    }
}

How to migrate previous old version to latest 2.6.0.1?

  • Remove PaddleConfig.Default.* settings because it's delted in 2.6.0.1

  • Add one of following config in 2nd parameter in PaddleOcrAll:

    • PaddleDevice.Openblas()
    • PaddleDevice.Mkldnn()
    • PaddleDevice.Onnx()
    • PaddleDevice.Gpu()
    • PaddleDevice.Gpu().And(PaddleDevice.TensorRt(...))

    image

How to migrate < 2.7.0 version to latest 2.7.0 when using LocalV3?

  • Uninstall NuGet package: Sdcb.PaddleOCR.Models.LocalV3
  • Install NuGet pakcage: Sdcb.PaddleOCR.Models.Local
  • Update namespaces from Sdcb.PaddleOCR.Models.LocalV3 to Sdcb.PaddleOCR.Models.Local
Product Compatible and additional computed target framework versions.
.NET net5.0 was computed.  net5.0-windows was computed.  net6.0 was computed.  net6.0-android was computed.  net6.0-ios was computed.  net6.0-maccatalyst was computed.  net6.0-macos was computed.  net6.0-tvos was computed.  net6.0-windows was computed.  net7.0 was computed.  net7.0-android was computed.  net7.0-ios was computed.  net7.0-maccatalyst was computed.  net7.0-macos was computed.  net7.0-tvos was computed.  net7.0-windows was computed.  net8.0 was computed.  net8.0-android was computed.  net8.0-browser was computed.  net8.0-ios was computed.  net8.0-maccatalyst was computed.  net8.0-macos was computed.  net8.0-tvos was computed.  net8.0-windows was computed. 
.NET Core netcoreapp2.0 was computed.  netcoreapp2.1 was computed.  netcoreapp2.2 was computed.  netcoreapp3.0 was computed.  netcoreapp3.1 was computed. 
.NET Standard netstandard2.0 is compatible.  netstandard2.1 was computed. 
.NET Framework net461 was computed.  net462 was computed.  net463 was computed.  net47 was computed.  net471 was computed.  net472 was computed.  net48 was computed.  net481 was computed. 
MonoAndroid monoandroid was computed. 
MonoMac monomac was computed. 
MonoTouch monotouch was computed. 
Tizen tizen40 was computed.  tizen60 was computed. 
Xamarin.iOS xamarinios was computed. 
Xamarin.Mac xamarinmac was computed. 
Xamarin.TVOS xamarintvos was computed. 
Xamarin.WatchOS xamarinwatchos was computed. 
Compatible target framework(s)
Included target framework(s) (in package)
Learn more about Target Frameworks and .NET Standard.

NuGet packages (8)

Showing the top 5 NuGet packages that depend on Sdcb.PaddleOCR:

Package Downloads
Sdcb.PaddleOCR.Models.Online The ID prefix of this package has been reserved for one of the owners of this package by NuGet.org.

Provides on-demand downloading of PaddleOCR models.

Wlkr.SafePaddleOCR

基于PaddleSharp.PaddleOCR设计的线程安全模板,示例: SafePaddleOCR safePaddleOCR = new SafePaddleOCR(); string imgPath = @"DimTechStudio-Logo.png"; var res = safePaddleOCR.Run(imgPath); Console.WriteLine($"res: {res.data.Text}");

HHO.LV.OCR

Library packed for OCR

Sdcb.PaddleOCR.Models.Local The ID prefix of this package has been reserved for one of the owners of this package by NuGet.org.

Known models for PaddleOCR

BotSharp.Plugin.PaddleSharp

Package Description

GitHub repositories (3)

Showing the top 3 popular GitHub repositories that depend on Sdcb.PaddleOCR:

Repository Stars
babalae/better-genshin-impact
🍨BetterGI · 更好的原神 - 自动拾取 | 自动剧情 | 全自动钓鱼(AI) | 全自动七圣召唤 | 自动伐木 | 快速传送 | 自动刷本 - UI Automation Testing Tools For Genshin Impact
SciSharp/BotSharp
The AI Agent Framework in .NET
sdcb/PaddleSharp
.NET/C# binding for Baidu paddle inference library and PaddleOCR
Version Downloads Last updated
2.7.0.1 771 1/15/2024
2.7.0 4,036 8/14/2023
2.7.0-preview.1 94 8/10/2023
2.6.0.6-preview.8 112 8/5/2023
2.6.0.6-preview.7 68 8/5/2023
2.6.0.6-preview.6 81 8/4/2023
2.6.0.6-preview.5 126 7/16/2023
2.6.0.6-preview.4 108 7/11/2023
2.6.0.6-preview.3 76 7/10/2023
2.6.0.6-preview.1 100 7/6/2023
2.6.0.5 4,430 6/17/2023
2.6.0.4 733 5/4/2023
2.6.0.3 453 5/3/2023
2.6.0.2 806 3/31/2023
2.6.0.1 2,031 12/8/2022
2.6.0 1,811 9/19/2022
2.6.0-preview6 203 9/10/2022
2.6.0-preview5 148 9/10/2022
2.6.0-preview4 147 8/27/2022
2.6.0-preview3 148 8/24/2022
2.5.0 3,469 8/1/2022
2.3.0 894 6/27/2022
2.2.2 1,973 2/18/2022