arabic-ocr

This research aims to fine-tune an Arabic OCR model using Tesseract 5.0, enhancing text recognition accuracy through extensive data collection, preprocessing, and image generation. By leveraging advanced training techniques and data augmentation, we achieve significant improvements in word error rates (WER).

In this repository, OCR-related datasets are available.

Arabic Chat with PDF is a user-friendly application that lets you interact with Arabic PDF documents. Powered by advanced language models, OCR, and vector search, it allows you to upload PDFs, ask questions, and receive accurate Arabic responses 🚀

الدليل الحي والمفتوح للبرمجيات والنماذج مفتوحة المصدر للغة العربية | The centralized living directory & showcase for Arabic open-source software, NLP, models, and tools.

Alef-OCR-Image2Html, an OCR model designed to transform Arabic documents including historical texts, scanned pages, and handwritten materials into structured and semantic HTML.

Automatic license-plate recognition for Sudanese plates — YOLO detector + fine-tuned OCR, with a reproducible benchmark.

The largest publicly released line-level dataset of historical Arabic manuscripts — 14 books, 3,043 pages, 28,600 lines, with margin/insertion-anchor annotations for non-linear reading order.

Official code for "Ketaba-OCR at AR-MS NakbaNLP 2026" — QLoRA fine-tuning of a specialized HTR model with Linear+Boost ensemble for Arabic manuscript recognition. 1st place per-line (CER 0.082) and 3rd place official leaderboard at NakbaNLP 2026 (LREC 2026).

Optical Character Recognition, OCR pipeline, Arabic OCR, Deep Learning OCR, Computer Vision text extraction, Text recognition system, AI document processing, Multilingual OCR, Transformer OCR, OCR benchmarking, Bounding box detection, Ground truth evaluation.

Fast, zero-dependency screen OCR to clipboard for Linux (Wayland & X11). Auto multi-language (Arabic + English).

Arabic handwritten text recognition using a CRNN (CNN + BiLSTM) with CTC loss, trained on the KHATT dataset — includes a Gradio web demo for OCR on your own images.

Additional experimental model for NakbaNLP 2026 Shared Task (AR-MS) — LoRA/DoRA fine-tuning of Qari-OCR (Qwen2-VL-2B) for Arabic handwritten manuscript recognition on the Omar Al-Saleh Memoir Collection (1951-1965).

Arabic OCR and diacritization pipeline with FastAPI backend, Android client, and reproducible notebook workflow.

Nassij V3: High-accuracy Arabic PDF-to-DOCX converter with direct digital extraction (NassijScanner) and cryptographic linguistic integrity verification (Merkle proofs).

DocuVision Showcase — commercial Arabic/English OCR desktop application for document review, search, and export; source code is not included.

DocuVision Electron — Arabic/English OCR desktop app for local document scanning, review, search, and export on Windows and Linux.

Reproducible Arabic OCR benchmark by Clouda OCR — distorted documents, normalized Arabic CER, model comparisons, methodology, and public audit artifacts.

Computer vision and Arabic OCR pipeline for structured information extraction from Egyptian National ID cards. YOLO card/field detection, EasyOCR Arabic text, per-field confidence and National ID validation.

Local, open-source OCR pipeline for Egyptian National ID cards — CLI, HTTP API, MCP server and Python SDK. No cloud APIs, nothing leaves your machine.

Official Clouda OCR website — Arabic-first OCR and Document AI, benchmark evidence, roadmap, support, and project information.