README.md

Web application for Sino-Nôm digitalization

Demo: https://share.streamlit.io/ds4v/nomnasite/main/app.py

Usage

pip install requirements.txt
streamlit run app.py

👉 Check out YouTube demo

Features

1. Input image from local or URL.
2. Leverage DL models to extract text from image:
   • Use VNPF's site as collected source.
   • Apply models based on the results of NomNaOCR.
3. Interactive mode using streamlit-drawable-canvas:
   • Drawing mode: draw rectangle boxes on image regions containing characters.
   • Editing mode: rotate, skew, scale, move any box of the canvas on demand.
   • Undo, Redo or Delete canvas contents.
4. Saving OCR results:
   • Export detection, recognition, and translation results to CSV or JSON.
   • Download patches cropped from detected bounding boxes.
5. Translate using APIs from:
   • VNUHCM University of Science: https://www.clc.hcmus.edu.vn/?page_id=3039
   • Sino-Nôm dictionary: https://hvdic.thivien.net/transcript.php#trans

(*) Note: In Editing mode, double-click a box to remove it.

Reference

My Vietnamese Sino-Nôm digitalization series :

• NomNaOCR: Optical Character Recognition.
• NomNaNMT: Neural Machine Translation.
• NomNaSite: Web Application.