Documentation
Academic-grade OCR engine for Mon script. High-performance, private, and localized.
1 Installation
Install the latest stable release of the MonOCR engine via our package manager. We recommend using a virtual environment for academic consistency.
pip install monocr
2 Quick Start Example
Initialize the processor and run text detection on a high-resolution scan.
# 1. Install engine
pip install monocr
# 2. Run OCR
from monocr import MonOCR
ocr = MonOCR()
text = ocr.predict('document_scan.png')
print(text)Input Standards
Follow these standards to achieve maximum recognition accuracy (97.5%+).
Resolution
Aim for a minimum of 300 DPI. For micro-text or ancient manuscripts, 600 DPI is required.
Lighting
Use diffuse lighting to minimize glares and deep shadows that can confuse the segmenter.
security Privacy-First OCR
Documents are processed entirely on your local machine. No data is uploaded to our servers.
- check_circle Localized data handling (GDPR/CCPA compliant).
- check_circle Opt-in, anonymous performance telemetry.
Technical Engine
Neural Architecture
MonOCR uses a neural-network architecture (MobileNetV3 + BiLSTM) optimized for archival digitization. Inference runs entirely in your browser using WebGPU/WASM for speed and privacy.
Preserving Mon Heritage
Mon (mnw) is a vulnerable language with limited digital presence. Most written knowledge remains locked in analog scans, making it difficult to build modern digital tools.
Your contributions directly help digitize this history. Every document or typed script improves our specialized AI model, enabling future research and linguistic preservation.
Model Hub (Hugging Face)
Our production weights and multi-format exports are hosted on Hugging Face for direct access.
Direct Access
You can download models directly for custom implementations or specialized deployment environments.
CLI Reference
Use the command line interface for processing large PDFs (>50MB) or batch folders.
Installation
pip install monocr
Basic Usage
monocr read image.png
monocr batch folder_path/
Multi-Platform SDKs
Official libraries for high-performance inference. Optimized for archival digitization and research.
Getting Started (Node.js)
// 1. Install
npm install monocr
// 2. Use
import { MonOCR } from 'monocr';
const ocr = new MonOCR();
const text = await ocr.predict('page.jpg');Installation (Python/CLI)
pip install monocr-onnx
Usage Guide
from monocr import MonOCR
ocr = MonOCR()
# 1. Read Image
text = ocr.predict("page.jpg")
# 2. Read PDF Document
pdf_text = ocr.predict("archival.pdf")
# 3. Extract Accuracy / Confidence
res = ocr.predict_with_confidence("sample.png")
print(f"Confidence: {res['confidence']:.2%}")Getting Started (Go)
// 1. Install
go get github.com/MonDevHub/monocr-onnx/go
// 2. Use
import "github.com/MonDevHub/monocr-onnx/go/pkg/ocr"
engine, _ := ocr.NewMonOCR("")
text, _ := engine.Predict("page.jpg")Getting Started (Rust)
// 1. Install (cargo add)
cargo add monocr-onnx
// 2. Use
use monocr_onnx::MonOCR;
let ocr = MonOCR::new("monocr.onnx")?;
let text = ocr.predict("page.jpg")?;License
MonOCR is released under the MIT License.