Parse Vision is an open-source tool for visualising what OCR is parsing in a PDF document. It helps developers and product teams identify if the parsing has missed vital information.
I was working on a proof of concept for an IDP platform (Intelligent Document Processing) when I noticed that some information was missing from documents. So, I decided to visualise the output to identify missing information quickly and validate which library would work better for different use cases.
Seeing many tutorials online claiming that parsing PDFs is easy and adding LLMs on top to extract information, I decided to open-source the tool to help developers and product teams make informed decisions about which library to use.
In the demo video, you'll notice that EasyOCR missed a quantity in the table, and Tesseract missed all the quantities.
- Clone the repo.
pip install -r requirements.txt
streamlit run index.py
You can try the PDF example in the example
folder.
Credit: docmosis.com
demo.mp4
The tool runs locally, no data are sent outside your machine, no external logs, and no telemetry.
- Tesseract
- EasyOCR
- Configure the level of detection (blocks, lines, words).
- Configure using GPU or CPU.
- Add more libraries, such as PaddleOCR
Feel free to connect with me on LinkedIn Oras Al-Kubaisi