如何使用 C# 增强图像质量以获得更好的 OCR 结果

// PaddleOCR: three model directories, all must exist and match the wrapper version
FullOcrModel models = new FullOcrModel(
    LocalDetectionModel.FromDirectory("models/ch_PP-OCRv4_det_infer"),       // ~5MB
    LocalClassificationModel.FromDirectory("models/ch_ppocr_mobile_v2.0_cls_infer"), // ~2MB
    LocalRecognitionModel.FromDirectory("models/ch_PP-OCRv4_rec_infer")      // ~15MB
);
using PaddleOcrAll ocr = new PaddleOcrAll(models);
using Mat mat = Cv2.ImRead("document.png");  // OpenCvSharp required for every image
PaddleOcrResult result = ocr.Run(mat);

Imports OpenCvSharp
Imports PaddleOCR

' PaddleOCR: three model directories, all must exist and match the wrapper version
Dim models As New FullOcrModel(
    LocalDetectionModel.FromDirectory("models/ch_PP-OCRv4_det_infer"),       ' ~5MB
    LocalClassificationModel.FromDirectory("models/ch_ppocr_mobile_v2.0_cls_infer"), ' ~2MB
    LocalRecognitionModel.FromDirectory("models/ch_PP-OCRv4_rec_infer")      ' ~15MB
)

Using ocr As New PaddleOcrAll(models)
    Using mat As Mat = Cv2.ImRead("document.png")  ' OpenCvSharp required for every image
        Dim result As PaddleOcrResult = ocr.Run(mat)
    End Using
End Using

// IronOCR: one package, zero model management
var ocr = new IronTesseract();
using var input = new OcrInput();
input.LoadImage("document.png");
var result = ocr.Read(input);
Console.WriteLine(result.Text);

Imports IronOcr

Dim ocr As New IronTesseract()
Using input As New OcrInput()
    input.LoadImage("document.png")
    Dim result = ocr.Read(input)
    Console.WriteLine(result.Text)
End Using

dotnet remove package Sdcb.PaddleOCR
dotnet remove package Sdcb.PaddleOCR.Models.Online
dotnet remove package Sdcb.PaddleInference.runtime.win64.mkl
dotnet remove package OpenCvSharp4
dotnet remove package OpenCvSharp4.runtime.win

dotnet remove package Sdcb.PaddleInference.runtime.win64.cuda118

// Before (PaddleOCR)
using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models;
using Sdcb.PaddleOCR.Models.Online;
using Sdcb.PaddleInference;
using OpenCvSharp;

// After (IronOCR)
using IronOcr;

Imports IronOcr
Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models
Imports Sdcb.PaddleOCR.Models.Online
Imports Sdcb.PaddleInference
Imports OpenCvSharp

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

// Local model configuration — developer owns the directory structure
// Each wrapper update may require re-downloading model files
string modelsRoot = Path.Combine(AppContext.BaseDirectory, "models");

FullOcrModel models = new FullOcrModel(
    LocalDetectionModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_PP-OCRv4_det_infer")),
    LocalClassificationModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_ppocr_mobile_v2.0_cls_infer")),
    LocalRecognitionModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_PP-OCRv4_rec_infer"))
);

// Fails at runtime if any of the three directories is missing or stale
using PaddleOcrAll ocr = new PaddleOcrAll(models)
{
    AllowRotateDetection = true,
    Enable180Classification = true
};

using Mat mat = Cv2.ImRead("document.png");
PaddleOcrResult result = ocr.Run(mat);
Console.WriteLine(result.Text);

Imports System
Imports System.IO

' Local model configuration — developer owns the directory structure
' Each wrapper update may require re-downloading model files
Dim modelsRoot As String = Path.Combine(AppContext.BaseDirectory, "models")

Dim models As New FullOcrModel(
    LocalDetectionModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_PP-OCRv4_det_infer")),
    LocalClassificationModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_ppocr_mobile_v2.0_cls_infer")),
    LocalRecognitionModel.FromDirectory(
        Path.Combine(modelsRoot, "ch_PP-OCRv4_rec_infer"))
)

' Fails at runtime if any of the three directories is missing or stale
Using ocr As New PaddleOcrAll(models) With {
    .AllowRotateDetection = True,
    .Enable180Classification = True
}

    Using mat As Mat = Cv2.ImRead("document.png")
        Dim result As PaddleOcrResult = ocr.Run(mat)
        Console.WriteLine(result.Text)
    End Using

End Using

// 否 model directories, no path configuration, no version matching
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

var ocr = new IronTesseract();

using var input = new OcrInput();
input.LoadImage("document.png");

var result = ocr.Read(input);
Console.WriteLine(result.Text);

Imports IronOcr

' 否 model directories, no path configuration, no version matching
IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

Dim ocr = New IronTesseract()

Using input As New OcrInput()
    input.LoadImage("document.png")

    Dim result = ocr.Read(input)
    Console.WriteLine(result.Text)
End Using

using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models.Online;
using OpenCvSharp;

// Separate async initialization step — blocks startup for 3-5 seconds on cold run
FullOcrModel models = await OnlineFullModels.EnglishV4.DownloadAsync();

using PaddleOcrAll ocr = new PaddleOcrAll(models)
{
    AllowRotateDetection = true,        // Enables 0/90/180/270 degree rotation detection
    Enable180Classification = true      // Additional pass for upside-down text
};

// OpenCV Mat required — no direct file path support
using Mat mat = Cv2.ImRead("rotated-scan.png");

if (mat.Empty())
{
    throw new FileNotFoundException("Image could not be loaded by OpenCvSharp");
}

// Three neural network passes: detection → classification → recognition
PaddleOcrResult result = ocr.Run(mat);

// Regions arrive in spatial order, not reading order
// Manual sort required for top-to-bottom, left-to-right output
var orderedRegions = result.Regions
    .OrderBy(r => r.Rect.Center.Y)
    .ThenBy(r => r.Rect.Center.X);

foreach (var region in orderedRegions)
{
    Console.WriteLine($"{region.Text} (confidence: {region.Score:P1})");
}

Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models.Online
Imports OpenCvSharp

' Separate async initialization step — blocks startup for 3-5 seconds on cold run
Dim models As FullOcrModel = Await OnlineFullModels.EnglishV4.DownloadAsync()

Using ocr As New PaddleOcrAll(models) With {
    .AllowRotateDetection = True,        ' Enables 0/90/180/270 degree rotation detection
    .Enable180Classification = True      ' Additional pass for upside-down text
}

    ' OpenCV Mat required — no direct file path support
    Using mat As Mat = Cv2.ImRead("rotated-scan.png")

        If mat.Empty() Then
            Throw New FileNotFoundException("Image could not be loaded by OpenCvSharp")
        End If

        ' Three neural network passes: detection → classification → recognition
        Dim result As PaddleOcrResult = ocr.Run(mat)

        ' Regions arrive in spatial order, not reading order
        ' Manual sort required for top-to-bottom, left-to-right output
        Dim orderedRegions = result.Regions _
            .OrderBy(Function(r) r.Rect.Center.Y) _
            .ThenBy(Function(r) r.Rect.Center.X)

        For Each region In orderedRegions
            Console.WriteLine($"{region.Text} (confidence: {region.Score:P1})")
        Next
    End Using
End Using

using IronOcr;

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

var ocr = new IronTesseract();

using var input = new OcrInput();
input.LoadImage("rotated-scan.png");
input.Deskew();    // Corrects rotation and skew automatically

var result = ocr.Read(input);

// Output is already in reading order — no sort needed
foreach (var page in result.Pages)
{
    foreach (var line in page.Lines)
    {
        Console.WriteLine($"{line.Text} (confidence: {line.Confidence}%)");
    }
}

Imports IronOcr

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

Dim ocr As New IronTesseract()

Using input As New OcrInput()
    input.LoadImage("rotated-scan.png")
    input.Deskew() ' Corrects rotation and skew automatically

    Dim result = ocr.Read(input)

    ' Output is already in reading order — no sort needed
    For Each page In result.Pages
        For Each line In page.Lines
            Console.WriteLine($"{line.Text} (confidence: {line.Confidence}%)")
        Next
    Next
End Using

using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models.Online;
using Sdcb.PaddleInference;    // GPU configuration namespace
using OpenCvSharp;

// Prerequisites must exist on every deployment environment:
// - NVIDIA Driver 452.39+ (Windows) / 450.80.02+ (Linux)
// - CUDA Toolkit 11.8 (not 12.x — version must match exactly)
// - cuDNN 8.6.0+ placed in CUDA bin directory
// - dotnet add package Sdcb.PaddleInference.runtime.win64.cuda118

FullOcrModel models = await OnlineFullModels.ChineseV4.DownloadAsync();

// GPU device 0, 1000MB initial memory pool
// Throws native load exception if CUDA_PATH not set or cuDNN DLL missing
using PaddleOcrAll ocr = new PaddleOcrAll(models, PaddleDevice.Gpu(deviceId: 0))
{
    AllowRotateDetection = true,
    Enable180Classification = true
};

using Mat mat = Cv2.ImRead("scanned-batch.png");
PaddleOcrResult result = ocr.Run(mat);

Console.WriteLine($"Text regions: {result.Regions.Length}");
Console.WriteLine(result.Text);

Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models.Online
Imports Sdcb.PaddleInference    ' GPU configuration namespace
Imports OpenCvSharp

' Prerequisites must exist on every deployment environment:
' - NVIDIA Driver 452.39+ (Windows) / 450.80.02+ (Linux)
' - CUDA Toolkit 11.8 (not 12.x — version must match exactly)
' - cuDNN 8.6.0+ placed in CUDA bin directory
' - dotnet add package Sdcb.PaddleInference.runtime.win64.cuda118

Dim models As FullOcrModel = Await OnlineFullModels.ChineseV4.DownloadAsync()

' GPU device 0, 1000MB initial memory pool
' Throws native load exception if CUDA_PATH not set or cuDNN DLL missing
Using ocr As New PaddleOcrAll(models, PaddleDevice.Gpu(deviceId:=0)) With {
    .AllowRotateDetection = True,
    .Enable180Classification = True
}

    Using mat As Mat = Cv2.ImRead("scanned-batch.png")
        Dim result As PaddleOcrResult = ocr.Run(mat)

        Console.WriteLine($"Text regions: {result.Regions.Length}")
        Console.WriteLine(result.Text)
    End Using
End Using

using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models.Online;
using OpenCvSharp;
using System.Collections.Generic;

FullOcrModel models = await OnlineFullModels.EnglishV4.DownloadAsync();
using PaddleOcrAll ocr = new PaddleOcrAll(models);
using Mat mat = Cv2.ImRead("invoice.png");

PaddleOcrResult result = ocr.Run(mat);

// 否 paragraph or line grouping — must implement manually
// Group regions into lines by proximity on the Y axis
var lineGroups = new Dictionary<int, List<PaddleOcrResultRegion>>();

foreach (var region in result.Regions)
{
    // Round Y center to nearest 15 pixels to approximate line grouping
    int lineKey = (int)(region.Rect.Center.Y / 15) * 15;

    if (!lineGroups.ContainsKey(lineKey))
        lineGroups[lineKey] = new List<PaddleOcrResultRegion>();

    lineGroups[lineKey].Add(region);
}

// Sort lines top to bottom, then regions left to right within each line
foreach (var line in lineGroups.OrderBy(kv => kv.Key))
{
    var lineText = string.Join(" ", line.Value
        .OrderBy(r => r.Rect.Center.X)
        .Select(r => r.Text));

    Console.WriteLine(lineText);
}

Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models.Online
Imports OpenCvSharp
Imports System.Collections.Generic

Dim models As FullOcrModel = Await OnlineFullModels.EnglishV4.DownloadAsync()
Using ocr As New PaddleOcrAll(models)
    Using mat As Mat = Cv2.ImRead("invoice.png")
        Dim result As PaddleOcrResult = ocr.Run(mat)

        ' 否 paragraph or line grouping — must implement manually
        ' Group regions into lines by proximity on the Y axis
        Dim lineGroups As New Dictionary(Of Integer, List(Of PaddleOcrResultRegion))()

        For Each region In result.Regions
            ' Round Y center to nearest 15 pixels to approximate line grouping
            Dim lineKey As Integer = CInt(region.Rect.Center.Y / 15) * 15

            If Not lineGroups.ContainsKey(lineKey) Then
                lineGroups(lineKey) = New List(Of PaddleOcrResultRegion)()
            End If

            lineGroups(lineKey).Add(region)
        Next

        ' Sort lines top to bottom, then regions left to right within each line
        For Each line In lineGroups.OrderBy(Function(kv) kv.Key)
            Dim lineText As String = String.Join(" ", line.Value _
                .OrderBy(Function(r) r.Rect.Center.X) _
                .Select(Function(r) r.Text))

            Console.WriteLine(lineText)
        Next
    End Using
End Using

using IronOcr;

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

var ocr = new IronTesseract();

using var input = new OcrInput();
input.LoadImage("invoice.png");

var result = ocr.Read(input);

// Hierarchical structure: Pages → Paragraphs → Lines → Words → Characters
// All delivered in reading order by the layout engine
foreach (var page in result.Pages)
{
    Console.WriteLine($"Page {page.PageNumber} — {page.Words.Count} words");

    foreach (var paragraph in page.Paragraphs)
    {
        Console.WriteLine($"  Paragraph at ({paragraph.X}, {paragraph.Y}):");
        Console.WriteLine($"  {paragraph.Text}");
    }
}

Imports IronOcr

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

Dim ocr As New IronTesseract()

Using input As New OcrInput()
    input.LoadImage("invoice.png")

    Dim result = ocr.Read(input)

    ' Hierarchical structure: Pages → Paragraphs → Lines → Words → Characters
    ' All delivered in reading order by the layout engine
    For Each page In result.Pages
        Console.WriteLine($"Page {page.PageNumber} — {page.Words.Count} words")

        For Each paragraph In page.Paragraphs
            Console.WriteLine($"  Paragraph at ({paragraph.X}, {paragraph.Y}):")
            Console.WriteLine($"  {paragraph.Text}")
        Next
    Next
End Using

using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models.Online;
using OpenCvSharp;
// Requires additional package: PdfSharp, iTextSharp, or similar
// Manual coordinate remapping from OpenCV pixel space to PDF point space

FullOcrModel models = await OnlineFullModels.EnglishV4.DownloadAsync();
using PaddleOcrAll ocr = new PaddleOcrAll(models);
using Mat mat = Cv2.ImRead("scanned-page.png");

PaddleOcrResult paddleResult = ocr.Run(mat);

// 否 built-in searchable PDF output — must build with external library
// region.Rect coordinates are in pixel space, PDF uses points (1 point = 1/72 inch)
// DPI conversion required for coordinate mapping
float dpiScale = 72.0f / 96.0f;  // Assuming 96 DPI source image

// ... hundreds of lines of PDF construction code using external library ...
// This is permanent maintenance, not a one-time cost
Console.WriteLine("Searchable PDF output requires external PDF library and coordinate mapping.");

Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models.Online
Imports OpenCvSharp
' Requires additional package: PdfSharp, iTextSharp, or similar
' Manual coordinate remapping from OpenCV pixel space to PDF point space

Dim models As FullOcrModel = Await OnlineFullModels.EnglishV4.DownloadAsync()
Using ocr As New PaddleOcrAll(models)
    Using mat As Mat = Cv2.ImRead("scanned-page.png")
        Dim paddleResult As PaddleOcrResult = ocr.Run(mat)

        ' 否 built-in searchable PDF output — must build with external library
        ' region.Rect coordinates are in pixel space, PDF uses points (1 point = 1/72 inch)
        ' DPI conversion required for coordinate mapping
        Dim dpiScale As Single = 72.0F / 96.0F  ' Assuming 96 DPI source image

        ' ... hundreds of lines of PDF construction code using external library ...
        ' This is permanent maintenance, not a one-time cost
        Console.WriteLine("Searchable PDF output requires external PDF library and coordinate mapping.")
    End Using
End Using

using IronOcr;

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

var ocr = new IronTesseract();

using var input = new OcrInput();
input.LoadImage("scanned-page.png");
input.Deskew();
input.DeNoise();

var result = ocr.Read(input);

// Searchable PDF in one line — no external PDF library, no coordinate mapping
result.SaveAsSearchablePdf("searchable-output.pdf");

Console.WriteLine($"Searchable PDF created. Confidence: {result.Confidence}%");

Imports IronOcr

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

Dim ocr As New IronTesseract()

Using input As New OcrInput()
    input.LoadImage("scanned-page.png")
    input.Deskew()
    input.DeNoise()

    Dim result = ocr.Read(input)

    ' Searchable PDF in one line — no external PDF library, no coordinate mapping
    result.SaveAsSearchablePdf("searchable-output.pdf")

    Console.WriteLine($"Searchable PDF created. Confidence: {result.Confidence}%")
End Using

using Sdcb.PaddleOCR;
using Sdcb.PaddleOCR.Models.Online;
using OpenCvSharp;
using System.Drawing;  // For multi-frame TIFF extraction
using System.Drawing.Imaging;
using System.Text;

FullOcrModel models = await OnlineFullModels.EnglishV4.DownloadAsync();
using PaddleOcrAll ocr = new PaddleOcrAll(models);

var fullText = new StringBuilder();

// Must use System.Drawing to extract individual TIFF frames
// OpenCvSharp cannot enumerate TIFF frames directly
using var tiff = Image.FromFile("multipage-scan.tiff");
FrameDimension dimension = new FrameDimension(tiff.FrameDimensionsList[0]);
int frameCount = tiff.GetFrameCount(dimension);

for (int i = 0; i < frameCount; i++)
{
    tiff.SelectActiveFrame(dimension, i);

    // Save frame to temp file — OpenCvSharp needs a file path
    string tempPath = Path.GetTempFileName() + ".png";
    tiff.Save(tempPath, ImageFormat.Png);

    try
    {
        using Mat mat = Cv2.ImRead(tempPath);
        PaddleOcrResult result = ocr.Run(mat);
        fullText.AppendLine($"=== Frame {i + 1} ===");
        fullText.AppendLine(result.Text);
    }
    finally
    {
        File.Delete(tempPath);  // Must clean up temp files
    }
}

Console.WriteLine(fullText.ToString());

Imports Sdcb.PaddleOCR
Imports Sdcb.PaddleOCR.Models.Online
Imports OpenCvSharp
Imports System.Drawing  ' For multi-frame TIFF extraction
Imports System.Drawing.Imaging
Imports System.Text

Dim models As FullOcrModel = Await OnlineFullModels.EnglishV4.DownloadAsync()
Using ocr As New PaddleOcrAll(models)
    Dim fullText As New StringBuilder()

    ' Must use System.Drawing to extract individual TIFF frames
    ' OpenCvSharp cannot enumerate TIFF frames directly
    Using tiff As Image = Image.FromFile("multipage-scan.tiff")
        Dim dimension As New FrameDimension(tiff.FrameDimensionsList(0))
        Dim frameCount As Integer = tiff.GetFrameCount(dimension)

        For i As Integer = 0 To frameCount - 1
            tiff.SelectActiveFrame(dimension, i)

            ' Save frame to temp file — OpenCvSharp needs a file path
            Dim tempPath As String = Path.GetTempFileName() & ".png"
            tiff.Save(tempPath, ImageFormat.Png)

            Try
                Using mat As Mat = Cv2.ImRead(tempPath)
                    Dim result As PaddleOcrResult = ocr.Run(mat)
                    fullText.AppendLine($"=== Frame {i + 1} ===")
                    fullText.AppendLine(result.Text)
                End Using
            Finally
                File.Delete(tempPath)  ' Must clean up temp files
            End Try
        Next
    End Using

    Console.WriteLine(fullText.ToString())
End Using

using IronOcr;

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY";

var ocr = new IronTesseract();

using var input = new OcrInput();
input.LoadImageFrames("multipage-scan.tiff");  // All frames in one call

var result = ocr.Read(input);

foreach (var page in result.Pages)
{
    Console.WriteLine($"=== Frame {page.PageNumber} ===");
    Console.WriteLine(page.Text);
}

// Optionally save the entire multi-frame result as searchable PDF
result.SaveAsSearchablePdf("multipage-searchable.pdf");

Imports IronOcr

IronOcr.License.LicenseKey = "YOUR-LICENSE-KEY"

Dim ocr As New IronTesseract()

Using input As New OcrInput()
    input.LoadImageFrames("multipage-scan.tiff")  ' All frames in one call

    Dim result = ocr.Read(input)

    For Each page In result.Pages
        Console.WriteLine($"=== Frame {page.PageNumber} ===")
        Console.WriteLine(page.Text)
    Next

    ' Optionally save the entire multi-frame result as searchable PDF
    result.SaveAsSearchablePdf("multipage-searchable.pdf")
End Using

// PaddleOCR: filter at 0.8
var highConfidence = result.Regions.Where(r => r.Score >= 0.8);

//IronOCRequivalent: filter at 80
var highConfidence = result.Pages
    .SelectMany(p => p.Words)
    .Where(w => w.Confidence >= 80);

' PaddleOCR: filter at 0.8
Dim highConfidence = result.Regions.Where(Function(r) r.Score >= 0.8)

' IronOCRequivalent: filter at 80
Dim highConfidence = result.Pages _
    .SelectMany(Function(p) p.Words) _
    .Where(Function(w) w.Confidence >= 80)

// PaddleOCR: manual sort required for reading order
var lines = result.Regions
    .OrderBy(r => r.Rect.Center.Y)
    .ThenBy(r => r.Rect.Center.X)
    .Select(r => r.Text);

// IronOCR: reading order is the default
var lines = result.Pages[0].Lines.Select(l => l.Text);

Imports System.Linq

' PaddleOCR: manual sort required for reading order
Dim lines = result.Regions _
    .OrderBy(Function(r) r.Rect.Center.Y) _
    .ThenBy(Function(r) r.Rect.Center.X) _
    .Select(Function(r) r.Text)

' IronOCR: reading order is the default
Dim lines = result.Pages(0).Lines.Select(Function(l) l.Text)

// PaddleOCR: stream → temp file → Mat → OCR
string tempPath = Path.GetTempFileName() + ".png";
using (var fs = File.Create(tempPath))
    await imageStream.CopyToAsync(fs);
using Mat mat = Cv2.ImRead(tempPath);
PaddleOcrResult result = ocr.Run(mat);
File.Delete(tempPath);

// IronOCR: stream → OCR (no temp file)
using var input = new OcrInput();
input.LoadImage(imageStream);
var result = ocr.Read(input);

Imports System.IO
Imports OpenCvSharp
Imports IronOcr

' PaddleOCR: stream → temp file → Mat → OCR
Dim tempPath As String = Path.GetTempFileName() & ".png"
Using fs = File.Create(tempPath)
    Await imageStream.CopyToAsync(fs)
End Using
Using mat As Mat = Cv2.ImRead(tempPath)
    Dim result As PaddleOcrResult = ocr.Run(mat)
End Using
File.Delete(tempPath)

' IronOCR: stream → OCR (no temp file)
Using input As New OcrInput()
    input.LoadImage(imageStream)
    Dim result = ocr.Read(input)
End Using

// PaddleOCR: async forced by model download
public async Task<string> ExtractTextAsync(string imagePath)
{
    FullOcrModel models = await OnlineFullModels.EnglishV4.DownloadAsync();
    using PaddleOcrAll ocr = new PaddleOcrAll(models);
    using Mat mat = Cv2.ImRead(imagePath);
    return ocr.Run(mat).Text;
}

// IronOCR: synchronous — no async required unless the caller needs it
public string ExtractText(string imagePath)
{
    var ocr = new IronTesseract();
    using var input = new OcrInput();
    input.LoadImage(imagePath);
    return ocr.Read(input).Text;
}

Imports System.Threading.Tasks

' PaddleOCR: async forced by model download
Public Async Function ExtractTextAsync(imagePath As String) As Task(Of String)
    Dim models As FullOcrModel = Await OnlineFullModels.EnglishV4.DownloadAsync()
    Using ocr As New PaddleOcrAll(models)
        Using mat As Mat = Cv2.ImRead(imagePath)
            Return ocr.Run(mat).Text
        End Using
    End Using
End Function

' IronOCR: synchronous — no async required unless the caller needs it
Public Function ExtractText(imagePath As String) As String
    Dim ocr As New IronTesseract()
    Using input As New OcrInput()
        input.LoadImage(imagePath)
        Return ocr.Read(input).Text
    End Using
End Function

# Remove from CI/CD pipeline:
# - name: Cache PaddleOCR models
#   uses: actions/cache@v3
#   with:
#     path: models/
#     key: paddleocr-models-${{env.PADDLEOCR_VERSION}}

#IronOCRonly needs standard NuGet caching:
- name: Cache NuGet packages
  uses: actions/cache@v3
  with:
    path: ~/.nuget/packages
    key: nuget-${{hashFiles('**/*.csproj')}}

# Find all PaddleOCR namespace imports
grep -rn "using Sdcb.PaddleOCR" --include="*.cs" .

# Find all OpenCvSharp imports (added as PaddleOCR dependency)
grep -rn "using OpenCvSharp" --include="*.cs" .

# Find all Mat usage patterns
grep -rn "Cv2\.ImRead\|new Mat\|Mat mat" --include="*.cs" .

# Find all async model download calls
grep -rn "DownloadAsync\|OnlineFullModels\|LocalDetectionModel" --include="*.cs" .

# Find all GPU device configuration
grep -rn "PaddleDevice\|EnableUseGpu\|cuda" --include="*.cs" .

# Find all result region access patterns
grep -rn "result\.Regions\|region\.Score\|region\.Rect" --include="*.cs" .

# Locate model directory references in configuration files
grep -rn "PP-OCRv4\|cls_infer\|det_infer\|rec_infer" --include="*.cs" --include="*.json" --include="*.yaml" .