L2C submission

Inspiration

Our team chose this project because it looked to be the most difficult and we were up to the challenge. We were also interested in the computer vision aspect.

What it does

Our project uses PyMuPDF in order to extract the text and drawing from the PDF plans and parses them in JSON. we are also using

How we built it

We built it with hard work, coffee and PyMuPDF and opensource LLM.

Challenges we ran into

We were initially not familiar with technical plans and the drawings were not in the same format between the different projects

Accomplishments that we're proud of

We are proud to have successfully generate JSON data for the plans as well as a PDF file containing the different components of the building.

What we learned

We learned the challenges associated with validating technical plans for construction and the need for a generic solution that could help prevent costly mistakes from plan and drawing mismatch.

What's next for L2C submission

If the drawings would not change from the different projects, a custom solution would work. Otherwise, our generic solution that extracts the text and feeds it to a local LLM or vision model is probably the way to go. I would recommand buying a very good computer to run better models of pay for an API key with a supplier that don't retail data.