Invizo - Arabic Handwritten OCR Solution

Overview

This project provides an end-to-end solution for recognizing Arabic handwritten, printed text, and Arabic numbers from images and documents on a given template document. The system efficiently detects and extracts text using State-of-the-Art Real-Scene Text Detection Trained on detecting Arabic Handwritten Text. Powered by a highly capable Recognition Model trained on the synthesized dataset and publicly available handwritten datasets.

The Application integrates text detection and recognition models, making it a practical solution for real-world OCR tasks for batch recognition of documents of the same type.

---

Limitations

• The background underline of the text must be empty or with light ink for appropriate results
• The Current OCR Engine is unable to recognize handwritten styles with more complex and cursive nature

---

Installation

🔹 Windows Binaries (Recommended)

Download and install the latest version using the Invizo-OCR Windows Installer:

Download Invizo-OCR Installer

Run the installer and follow the setup instructions.

🔹 Manual Installation (Source Code)

For setting up the project from source, ensure your machine environment includes OpenCV Java .dll and follow these steps:

1. Clone the Repository

   git clone [https://github.com/yourusername/arabic-ocr.git](https://github.com/Hedrax/Invizo-OCR)
   cd Invizo-OCR

2. Set Up JavaFX and OpenCV

• Ensure you have Java 11+ installed.
• Download and set up the JavaFX SDK (Download).
• Configure your IDE (IntelliJ IDEA, Eclipse, or VS Code) to include JavaFX libraries.
• Add OpenCV Java .dll to your system's library path.

3. Build and Run the Application

   ./gradlew run

4. Once you run the application, the latest release's recognition and detection model will be downloaded automatically.

Note: If using Linux or Mac you will be required to extract the binary file from the python script in \src\main\python\ and change the operation connection in \src\main\java\com\example\ocrdesktop\data\LocalAiService.java

---

For More information about the approach and performance, check out the paper

Citation

@misc{waly2025invizoarabichandwrittendocument,
      title={Invizo: Arabic Handwritten Document Optical Character Recognition Solution}, 
      author={Alhossien Waly and Bassant Tarek and Ali Feteha and Rewan Yehia and Gasser Amr and Walid Gomaa and Ahmed Fares},
      year={2025},
      eprint={2502.05277},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2502.05277}, 
}

Demo Video

Google Drive: Invizo Demo Video

Screenshots

Receipt Template

Sheets

Document Type

All Requests

Detail Request 1

Detail Request 2

Detail Document

New Request