Project Description
I need a browser-based AI app that behaves much like bild.ai but tailored to doors. The user will upload a floor plan as a single PDF; the system must automatically read the drawing, identify every door, count them, and then draw on our own product universe to suggest the most suitable models in line with the client’s stated specifications.
The core workflow is simple: PDF in, AI vision and machine-learning routines do the heavy lifting, and a clear list of recommended door products comes out. I will supply the full catalogue, spec sheets and any training data you need so the suggestion engine can learn our range and stay current as it evolves.
Key features to build are automatic floor-plan parsing, door count analysis and a recommendation engine that maps detected door types to our catalogue. Please ensure the solution is entirely web based, no native installs. I expect a clean, intuitive interface for both end-users and an admin area where I can update the product database and retrain models as required.
Deliverables:
• Working web app deployed to a test server
• Trained model with documented inference pipeline
• Admin panel for catalogue maintenance
• Source code with setup instructions and brief technical doc
Acceptance will be based on accurate door counts from typical architectural PDFs and relevant product suggestions drawn only from our supplied catalogue.