Simple-LaTeX-OCR

The ecoder is resetv2+simplevit, the decoder is trasformer, the attetio mechaism uses flashattetio2.0 from pytorch2.0, ad the dataset size is 112 millio

Olie experiece

https://huggigface.co/spaces/chaodreamig/Simple-LaTeX-OCR

https://www.modelscope.c/studios/chaodreamig/Simple-LaTeX-OCR/summary

Istall the package simple_latex_ocr:

pip istall simple_latex_ocr

Use from withi Pytho

from simple_latex_ocr.models import Latex_OCR
model = Latex_OCR()
img_path = "tests/test_files/5.pg"
result = model.predict(img_path)
prit(result['formula'])
prit(result['cofidece'])
prit(result['elapse'])

Used by commad lie

$ simple_latex_ocr tests/test_files/2.pg

Used by api

$ pytho -m simple_latex_ocr.api.ru

#You ca use test.py to iitiate a request for verificatio

Cotributio

Cotributios of ay kid are welcome.

Ackowledgmet

Code take ad modified from lukas-blecher, RapidAI,ultralytics

Simple-LaTeX-OCR The encoder is resnetv2+simplevit, the decoder is transformer, the attention mechan