RustODotnet 0.2.0

RustODotnet

High-performance cross-platform OCR library for .NET based on RustO! and PaddleOCR, powered by the lightweight MNN inference engine.

📦 Installation

Install the core runtime package via NuGet:

dotnet add package RustODotnet

Choose an OCR Model Package

Models are distributed as modular NuGet packages so you can pick the tier that suits your app:

# Recommended default (~6 MB)
dotnet add package RustODotnet.Models.PPOCRv6.Tiny

# Higher accuracy PP-OCRv6 tiers
dotnet add package RustODotnet.Models.PPOCRv6.Small    # ~30 MB
dotnet add package RustODotnet.Models.PPOCRv6.Medium   # ~134 MB

# PP-OCRv5 tiers
dotnet add package RustODotnet.Models.PPOCRv5.Mobile   # ~28 MB
dotnet add package RustODotnet.Models.PPOCRv5.Server   # ~270 MB

# PP-OCRv4 tiers (with text orientation classifier)
dotnet add package RustODotnet.Models.PPOCRv4.Mobile   # ~23 MB
dotnet add package RustODotnet.Models.PPOCRv4.Server   # ~300 MB

Note: When referencing any model package, model files are automatically copied to your output models/ directory at build time.

🚀 Quick Start

1. Basic OCR (Automatic Model Discovery)

using System;
using RustODotnet;

// Initialize engine (automatically discovers models in models/ folder)
using var ocr = new RustO();

// Run OCR on an image file
var results = ocr.RecognizeFile("document.jpg");

foreach (var res in results)
{
    Console.WriteLine($"Text: '{res.Text}' (Confidence: {res.Score:P1})");
    Console.WriteLine($"  Frame: X={res.Frame.Left}, Y={res.Frame.Top}, W={res.Frame.Width}, H={res.Frame.Height}");
}

2. Formatted Spatial Layout Output

Preserve the spatial structure (columns, paragraphs, tables) of documents:

using var ocr = new RustO();

// Get text formatted as it appears visually
string spatialText = ocr.RecognizeFileToSpatialText("invoice.png");
Console.WriteLine(spatialText);

3. Advanced Configuration

using RustODotnet;

var config = new RustOConfig
{
    Template = "ppv6",
    DetModelPath = "models/det.mnn",
    RecModelPath = "models/rec.mnn",
    DictPath = "models/dict.txt",
    TextScore = 0.5f,
    DetThresh = 0.3f,
    DetBoxThresh = 0.6f,
    LimitSideLen = 960,
    UseDilation = true
};

using var ocr = new RustO(config);
var results = ocr.RecognizeFile("photo.jpg");

4. Recognize from In-Memory Bytes

byte[] imageBytes = File.ReadAllBytes("receipt.jpg");

using var ocr = new RustO();
var results = ocr.RecognizeBytes(imageBytes);

🌐 Supported Platforms

Native shared binaries are included out of the box for:

• Windows: x64 (rusto.dll)
• Linux: x64 (librusto.so)
• macOS: x64 & ARM64 Apple Silicon (librusto.dylib)

Target frameworks: net8.0, net6.0, netstandard2.1.

📄 License

MIT License.