This project implements a web app where users upload a photo/scan of a hand-drawn software/system/process diagram. The app extracts shapes, arrows, and text, builds a structured graph, and exports a JSON file representing the graphical elements.
The pipeline uses robust classical Computer Vision (OpenCV) plus OCR (Tesseract). An optional neural segmentation model scaffold is included for improved results (PyTorch).
- Install Tesseract.
- Linux:
sudo apt install tesseract-ocr - MacOS:
brew install tesseract - Windows: install from the internet and add to PATH.
- Install all required packages.
pip install -r requirements.txt
- Linux/MacOS:
bash run.sh - Windows:
run_windows.bat