How to Batch Image OCR with Generative AI API Calls

Env config

install uv and run the following command to install the dependencies:

uv sync

Run the demo

Prepare your own OpenAI API key, write it in .env. The run the following command in terminal:

• input_dir: the directory of your input images, e.g., sample_images
• collection_series: the collection series you want to use legacy (Generative AI Challenge) or sydow
• save-json: whether to save the OCR results in JSON format
• save-excel: whether to save the OCR results in Excel format

uv run python -m main --input-dir sample_images --collection-series legacy --save-json --save-excel

Ajust it to your project

You need to create a structured output Pydantic model in main.py and adjust batch_ocr function accoding, then change your promptsin ocr_prompt_default.yml to your needs.
You can read more details in FungariumOCR.pdf.
Enjoy :)