web-arabic-ocr

Simple AI-generated web interface for Arabic OCR using a vision-to-sequence model.

This repository provides a small Flask app that accepts pasted or uploaded images and returns OCR results for Arabic text.

Requirements

• Python 3.10 or 3.11
• Conda (Miniconda or Anaconda)

Setup

1. Open a terminal in the arab-ocr-flask directory.

2. Create the conda environment defined in environment.yml and activate it:

conda env create -f environment.yml
conda activate arab-ocr

3. Install PyTorch with CUDA support. The provided setup_env.sh installs the latest available pytorch-cuda via conda. Run it with:

bash setup_env.sh

Running the app

Start the Flask app from the project folder:

python app.py

By default the server runs in debug mode on http://127.0.0.1:5000/.

Notes and troubleshooting

• Model download: the first run may download large model files. Ensure you have sufficient disk space and a reliable connection.
• GPU drivers: installing the latest pytorch-cuda via conda usually selects a recent CUDA build. If CUDA is not compatible with your system drivers, install a specific pytorch-cuda version (for example pytorch-cuda=11.8).
• Memory: running large models requires significant RAM and GPU memory. If you run out of memory, use a smaller model or run on CPU.