Formula OCR

Steps to follow:

Using general methodology:

1. Clone the repo into formula_ocr_main directory:

    git clone https://github.com/Trojan0101/formula_ocr.git

2. Create a directory for file downloading:

    cd formula_ocr
    mkdir downloaded_images

3. Move tesseract trained datasets for english, japanese, korean, chinese traditional, and chinese simplified to tessdata path in server:

   • Use WinScp to move files to the server from https://github.com/tesseract-ocr/tessdata/tree/main

4. Install dependencies:

   pip install virtualenv

   virtualenv formula_ocr_env

   source formula_ocr_env/bin/activate

   pip install -r ./requirements.txt

   pip install uwsgi

5. Move modified utils.py to rapid_latex_ocr utils.py file:

    mv modified_site_packages/rapid_latex_ocr/utils.py formula_ocr_env/<python_version>/site_packages/rapid_latex_ocr/utils.py

6. Install missing libraries:

   sudo apt-get install libgl1-mesa-glx

7. Run:

   nohup uwsgi --http :8080 --module app:app > formula_ocr_main.log 2>&1 &

   disown

Using docker:

1. Load the Docker Image from the Tar File:

   docker load -i docker_images/formula_ocr_docker_linux.tar

2. Verify the Image is Loaded:

   docker images

3. Run a Container from the Image [Create formula_ocr_log directory]:

   docker run -p 8080:8080 -v /formula_ocr_log:/formula_ocr_docker --name ocr formula_ocr_docker

Notes:

1. Path to tessdata can be /usr/local/share/tessdata or /usr/share/tessdata