document-ocr

Layout preserving OCR for documents. Includes text, tables and figures. Useful for LEAP OCR and Bhashini apps API call.

Step 1 : Create Virtual Environment

Make sure you are using Python 3.10 and create a virtual environment to install upcoming dependencies

python3 -m venv <myenvpath>

Step 2 : Install Requirements

Use this virtual environment to install the following dependencies

pip install -r requirements.txt

Step 3 : Download Models

From the release section download the two models. Place figure-detector model in 'figures/model' and place sprint.pt for table strcuture recogniiton in 'tables/model' directory

Step 4 : Run the pipeline

Use main.py to set the input file parameters, output set name, language, table, and figures flag and execute as follows.

python3 main.py

Step 5 : Using the UI

You can also use the streamlit UI to execute the pipeline and download the compressed output.

streamlit run app.py