Hi community,
I’d like to share my project I just finished recently: Texo, an in-browser LaTeX OCR model built on Transformers and Transformers.js, which is a free and open-sourced alternative to Mathpix or SimpleTex.
It uses a lite but accurate model(only 20M parameters!) I finetuned by Transformers and ships with the Transformers.js framework. Hope this would help the STEM/AI learners to taking notes with LaTeX formula.
Training codes: GitHub - alephpi/Texo: A minimalist SOTA LaTeX OCR model which contains only 20M parameters
Front end: GitHub - alephpi/Texo-web: In-browser LaTeX formula OCR tool
Online demo: https://texocr.netlify.app/
Everything runs in your browser, no server, no deployment, no env configs compared to other famous LaTeX OCR open-source projects, you only need to wait for ~80MB model download from HF Hub at your first visit.
1 Like