Switch to Tesseract package

AGENTS.md

@@ -27,14 +27,7 @@
 - Use `Serilog__MinimumLevel=Verbose` to enable detailed timings and counts.
 
 ## Operations
-Le librerie native minime per Linux `x64` sono disponibili in `src/MarkItDownNet/TesseractOCR/x64` e vengono copiate accanto ai binari (`x64`) ad eccezione di `libdl.so`, posizionata in `runtimes/linux-x64/native`:
-
-* `libopenjp2.so.7`
-* `liblept.so.5` con il symlink `libleptonica-1.85.0.dll.so`
-* `libtesseract.so.5` con il symlink `libtesseract55.dll.so`
-* `libdl.so`
-
-Grazie a queste dipendenze la libreria è auto‑consistente e **non richiede l'installazione di Tesseract o Leptonica**.
+Le dipendenze native minime per Linux `x64` (Tesseract e Leptonica) sono incluse nel repository in `runtimes/linux-x64/native` e vengono copiate accanto ai binari. Non è richiesta l'installazione separata di Tesseract.
 
 Per l'OCR servono solo i dati delle lingue. Su Ubuntu 24.04 possono essere installati con:
 

README.md

@@ -46,14 +46,7 @@ All build and test commands must use the locally installed `dotnet`:
 
 ## Tesseract and leptonica
 
-Per l'esecuzione su Linux `x64` il progetto include solo le librerie native strettamente necessarie. Tesseract e Leptonica risiedono nella sottocartella `x64` accanto ai binari, mentre `libdl.so` è collocata sotto `runtimes/linux-x64/native` per soddisfare il loader di `TesseractOCR`:
-
-* `libopenjp2.so.7`
-* `liblept.so.5` e il symlink `libleptonica-1.85.0.dll.so`
-* `libtesseract.so.5` e il symlink `libtesseract55.dll.so`
-* `libdl.so`
-
-Grazie a queste dipendenze pre‑caricate la libreria è *auto‑consistente* e **non richiede l'installazione di Tesseract o Leptonica sul sistema**.
+La libreria include le dipendenze native minime per Linux `x64` in `runtimes/linux-x64/native` e non richiede l'installazione di Tesseract o Leptonica sul sistema.
 
 Per eseguire l'OCR è necessario soltanto fornire i file `tessdata` delle lingue. Su Ubuntu 24.04 è sufficiente installare i pacchetti delle lingue desiderate, ad esempio:
 

docs/funsd_comparison.md

@@ -61,6 +61,5 @@ Confronto tra le bounding box delle parole annotate nel dataset di test FUNSD e
 
 ```bash
 export PATH=$HOME/.dotnet:$PATH
-export LD_LIBRARY_PATH=/usr/lib/x86_64-linux-gnu:$PWD/tools/FunsdComparison/bin/Debug/net9.0/runtimes/linux-x64/native
 dotnet run --project tools/FunsdComparison
 ```

src/MarkItDownNet/MarkItDownConverter.cs

@@ -2,14 +2,11 @@
 using System.Collections.Generic;
 using System.IO;
 using System.Linq;
-using System.Runtime.InteropServices;
 using System.Threading;
 using System.Threading.Tasks;
 using Markdig;
 using Serilog;
-using TesseractOCR;
-using TesseractOCR.Enums;
-using TesseractOCR.InteropDotNet;
+using Tesseract;
 using UglyToad.PdfPig;
 using UglyToad.PdfPig.Content;
 using PDFtoImage;
@@ -23,40 +20,6 @@ public class MarkItDownConverter
     private readonly MarkItDownOptions _options;
     private readonly ILogger _logger;
 
-    static MarkItDownConverter()
-    {
-        if (RuntimeInformation.IsOSPlatform(OSPlatform.Linux))
-        {
-            var libDir = Path.Combine(AppContext.BaseDirectory, "x64");
-            var ldPath = Environment.GetEnvironmentVariable("LD_LIBRARY_PATH");
-            Environment.SetEnvironmentVariable("LD_LIBRARY_PATH", string.IsNullOrEmpty(ldPath) ? libDir : libDir + ":" + ldPath);
-            LibraryLoader.Instance.CustomSearchPath = libDir;
-            LoadNative("libopenjp2.so.7");
-            LoadNative("libleptonica-1.85.0.dll.so");
-            LoadNative("libtesseract.so.5");
-        }
-    }
-
-    private const int RTLD_NOW = 2;
-    private const int RTLD_GLOBAL = 0x100;
-
-    [DllImport("libdl.so.2")]
-    private static extern IntPtr dlopen(string fileName, int flags);
-
-    private static void LoadNative(string name)
-    {
-        var path = Path.Combine(AppContext.BaseDirectory, "x64", name);
-        if (!File.Exists(path))
-        {
-            throw new DllNotFoundException($"Unable to find '{name}' at '{path}'");
-        }
-        var handle = dlopen(path, RTLD_NOW | RTLD_GLOBAL);
-        if (handle == IntPtr.Zero)
-        {
-            throw new DllNotFoundException($"Unable to load '{name}' from '{path}'");
-        }
-    }
-
     public MarkItDownConverter(MarkItDownOptions? options = null, ILogger? logger = null)
     {
         _options = options ?? new MarkItDownOptions();
@@ -136,7 +99,7 @@ private MarkItDownResult ProcessPdfWithOcr(string path, CancellationToken ct)
                 pages.Add(new Page(pages.Count + 1, bitmap.Width, bitmap.Height));
                 using var image = SKImage.FromBitmap(bitmap);
                 using var data = image.Encode(SKEncodedImageFormat.Png, 100);
-                using var pix = TesseractOCR.Pix.Image.LoadFromMemory(data.ToArray());
+                using var pix = Pix.LoadFromMemory(data.ToArray());
                 var result = ProcessPix(pix, pages.Count, ct);
                 lines.AddRange(result.lines);
                 words.AddRange(result.words);
@@ -149,51 +112,44 @@ private MarkItDownResult ProcessPdfWithOcr(string path, CancellationToken ct)
 
     private MarkItDownResult ProcessImage(string path, CancellationToken ct)
     {
-        using var pix = TesseractOCR.Pix.Image.LoadFromFile(path);
+        using var pix = Pix.LoadFromFile(path);
         var (lines, words) = ProcessPix(pix, 1, ct);
         var pages = new List<Page> { new Page(1, pix.Width, pix.Height) };
         var markdown = BuildMarkdown(lines);
         return new MarkItDownResult(markdown, pages, lines, words);
     }
 
-    private (List<Line> lines, List<Word> words) ProcessPix(TesseractOCR.Pix.Image pix, int pageNumber, CancellationToken ct)
+    private (List<Line> lines, List<Word> words) ProcessPix(Pix pix, int pageNumber, CancellationToken ct)
     {
         var lines = new List<Line>();
         var words = new List<Word>();
-        using var engine = new Engine(_options.OcrDataPath ?? string.Empty, _options.OcrLanguages, EngineMode.Default);
+        using var engine = new TesseractEngine(_options.OcrDataPath ?? string.Empty, _options.OcrLanguages, EngineMode.Default);
         using var page = engine.Process(pix);
-
-        foreach (var block in page.Layout)
+        using var iter = page.GetIterator();
+        iter.Begin();
+        do
         {
-            foreach (var paragraph in block.Paragraphs)
+            ct.ThrowIfCancellationRequested();
+
+            if (iter.IsAtBeginningOf(PageIteratorLevel.TextLine) &&
+                iter.TryGetBoundingBox(PageIteratorLevel.TextLine, out var rectLine))
             {
-                foreach (var textLine in paragraph.TextLines)
+                var text = iter.GetText(PageIteratorLevel.TextLine)?.Trim() ?? string.Empty;
+                if (!string.IsNullOrEmpty(text))
                 {
-                    ct.ThrowIfCancellationRequested();
-                    if (textLine.BoundingBox is Rect rectLine)
-                    {
-                        var text = textLine.Text?.Trim() ?? string.Empty;
-                        if (!string.IsNullOrEmpty(text))
-                        {
-                            lines.Add(new Line(pageNumber, text, Normalize(rectLine, pix.Width, pix.Height)));
-                        }
-                    }
-
-                    foreach (var word in textLine.Words)
-                    {
-                        ct.ThrowIfCancellationRequested();
-                        if (word.BoundingBox is Rect rectWord)
-                        {
-                            var wText = word.Text?.Trim() ?? string.Empty;
-                            if (!string.IsNullOrEmpty(wText))
-                            {
-                                words.Add(new Word(pageNumber, wText, Normalize(rectWord, pix.Width, pix.Height)));
-                            }
-                        }
-                    }
+                    lines.Add(new Line(pageNumber, text, Normalize(rectLine, pix.Width, pix.Height)));
                 }
             }
-        }
+
+            if (iter.TryGetBoundingBox(PageIteratorLevel.Word, out var rectWord))
+            {
+                var wText = iter.GetText(PageIteratorLevel.Word)?.Trim() ?? string.Empty;
+                if (!string.IsNullOrEmpty(wText))
+                {
+                    words.Add(new Word(pageNumber, wText, Normalize(rectWord, pix.Width, pix.Height)));
+                }
+            }
+        } while (iter.Next(PageIteratorLevel.Word));
 
         return (lines, words);
     }

src/MarkItDownNet/MarkItDownNet.csproj

@@ -13,36 +13,13 @@
     <PackageReference Include="Serilog.Settings.Configuration" Version="9.0.0" />
     <PackageReference Include="Serilog.Sinks.Console" Version="6.0.0" />
     <PackageReference Include="Serilog.Sinks.File" Version="7.0.0" />
-    <PackageReference Include="TesseractOCR" Version="5.5.1" />
+    <PackageReference Include="Tesseract" Version="5.2.0" />
     <PackageReference Include="PdfPig" Version="0.1.11" />
   </ItemGroup>
 
   <ItemGroup>
     <!-- Bundled native libraries for Linux (x64) -->
-    <None Include="TesseractOCR/x64/libtesseract.so.5">
-      <Link>x64/libtesseract.so.5</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
-    <None Include="TesseractOCR/x64/libtesseract55.dll.so">
-      <Link>x64/libtesseract55.dll.so</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
-    <None Include="TesseractOCR/x64/liblept.so.5">
-      <Link>x64/liblept.so.5</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
-    <None Include="TesseractOCR/x64/libleptonica-1.85.0.dll.so">
-      <Link>x64/libleptonica-1.85.0.dll.so</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
-    <None Include="TesseractOCR/x64/libopenjp2.so.7">
-      <Link>x64/libopenjp2.so.7</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
-    <None Include="TesseractOCR/x64/libdl.so">
-      <Link>runtimes/linux-x64/native/libdl.so</Link>
-      <CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
-    </None>
+    <None Include="runtimes/linux-x64/native/*" CopyToOutputDirectory="PreserveNewest" Pack="true" PackagePath="runtimes/linux-x64/native" />
   </ItemGroup>
 
 </Project>

tests/MarkItDownNet.Tests/MarkItDownNet.Tests.csproj

@@ -14,7 +14,7 @@
     <PackageReference Include="Microsoft.NET.Test.Sdk" Version="17.14.1" />
     <PackageReference Include="xunit" Version="2.9.3" />
     <PackageReference Include="xunit.runner.visualstudio" Version="3.1.3" />
-    <PackageReference Include="TesseractOCR" Version="5.5.1" />
+    <PackageReference Include="Tesseract" Version="5.2.0" />
     <PackageReference Include="SkiaSharp" Version="3.119.0" />
     <PackageReference Include="SkiaSharp.NativeAssets.Linux" Version="3.119.0" />
   </ItemGroup>

tests/MarkItDownNet.Tests/OcrImageTests.cs

@@ -11,15 +11,6 @@ public class OcrImageTests
     [Fact]
     public async Task Can_extract_text_from_simple_png()
     {
-        try
-        {
-            OcrTestHelpers.EnsureOcrLibraries();
-        }
-        catch (Exception)
-        {
-            return;
-        }
-
         using var surface = SKSurface.Create(new SKImageInfo(120, 40));
         var canvas = surface.Canvas;
         canvas.Clear(SKColors.White);

tests/MarkItDownNet.Tests/OcrPdfTests.cs

@@ -5,8 +5,6 @@
 using System.Text.Json;
 using System.Threading.Tasks;
 using MarkItDownNet;
-using TesseractOCR;
-using TesseractOCR.InteropDotNet;
 using Xunit;
 
 namespace MarkItDownNet.Tests;
@@ -16,11 +14,6 @@ public class OcrPdfTests
     [Fact]
     public async Task OcrTestPdfMatchesGroundTruth()
     {
-        try {
-        OcrTestHelpers.EnsureOcrLibraries();
-    } catch (Exception) {
-        return;
-    }
         using var http = new HttpClient();
         var baseUrl = "https://raw.githubusercontent.com/docling-project/docling/main/tests/data_scanned";
 

tools/DoclingComparison/DoclingComparison.csproj

@@ -9,7 +9,7 @@
 
   <ItemGroup>
     <ProjectReference Include="../../src/MarkItDownNet/MarkItDownNet.csproj" />
-    <PackageReference Include="TesseractOCR" Version="5.5.1" />
+    <PackageReference Include="Tesseract" Version="5.2.0" />
     <PackageReference Include="BitMiracle.LibTiff.NET" Version="2.4.660" />
     <PackageReference Include="SkiaSharp" Version="3.119.0" />
     <PackageReference Include="SkiaSharp.NativeAssets.Linux" Version="3.119.0" />

tools/DoclingComparison/Program.cs

@@ -3,7 +3,6 @@
 using SkiaSharp;
 using System.Text;
 using System.Text.Json;
-using TesseractOCR.InteropDotNet;
 
 namespace DoclingComparison;
 
@@ -19,7 +18,6 @@ static async Task Main(string[] args)
         var groundRoot = Path.Combine(dataRoot, "groundtruth", "docling_v2");
         var reportPath = Path.Combine(repoRoot, "docs", "docling_comparison.md");
 
-        LibraryLoader.Instance.CustomSearchPath = "/lib/x86_64-linux-gnu";
         Environment.SetEnvironmentVariable("TESSDATA_PREFIX", "/usr/share/tesseract-ocr/5/tessdata");
 
         var files = args.Length > 0 ? args : Directory.GetFiles(Path.Combine(dataRoot, "pdf"), "*.pdf")

tools/DoclingTimings/DoclingTimings.csproj

@@ -7,6 +7,6 @@
   </PropertyGroup>
   <ItemGroup>
     <ProjectReference Include="../../src/MarkItDownNet/MarkItDownNet.csproj" />
-    <PackageReference Include="TesseractOCR" Version="5.5.1" />
+    <PackageReference Include="Tesseract" Version="5.2.0" />
   </ItemGroup>
 </Project>