Skip to content

Latest commit

 

History

History
65 lines (45 loc) · 1.17 KB

README.md

File metadata and controls

65 lines (45 loc) · 1.17 KB

업스테이지 수학 수식 OCR 모델

Requirements

All dependencies can be installed with PIP.

pip install tensorboardX tqdm pyyaml psutil

현재 검증된 GPU 개발환경으로는

  • Pytorch 1.0.0 (CUDA 10.1)
  • Pytorch 1.4.0 (CUDA 10.0)
  • Pytorch 1.7.1 (CUDA 11.0)

Supported Models

Supported Data

  • Aida (synthetic handwritten)
  • CROHME (online handwritten)
  • IM2LATEX (pdf, synthetic handwritten)
  • Upstage (print, handwritten)

모든 데이터는 팀 저장소에서 train-ready 포맷으로 다운 가능하다.

[dataset]/
├── gt.txt
├── tokens.txt
└── images/
    ├── *.jpg
    ├── ...     
    └── *.jpg

Usage

Training

python train.py

Evaluation

python evaluate.py