CoContractor

Inspiration

Converting paper or PDF floor plans into 3D models and getting material estimates is a slow, manual process. Finding exact square footage for paint, tile, and countertops often takes days of site visits and calls with local suppliers. We built CoContractor to solve this—turning static 2D blueprints into interactive 3D models and automated, cost-optimized material lists in seconds.

What it does

CoContractor allows contractors and home flippers to upload a photo or PDF of any house blueprint. The app automatically:

Extracts exact room dimensions, surface areas, and wall heights.
Generates an interactive, web-based 3D model of the home layout.
Calculates required quantities of renovation materials like countertops, floor tiles, paint, and fixtures.
Queries nearby suppliers and ranks materials from lowest to highest cost to maximize project margins.

How we built it

CoContractor is powered by Microsoft Azure cloud services:

1) Blueprint Ingestion: Users upload a floor plan image. Azure AI Vision scans it to detect walls, doors, windows, and scale markers. 2) Dimension Extraction: The backend uses basic geometry to calculate exact surface areas and room volumes. 3) 3D Visualization: Extracted dimensions are passed to a Three.js web renderer hosted on Azure App Service, instantly building a navigable 3D model. 4) Material Optimizer: Azure OpenAI queries local supplier catalogs stored in Azure AI Search to match room dimensions with real-time costs, ordered from lowest to highest price.

Challenges we ran into

Messy or Low-Res Blueprints: Variations in image quality made wall detection tricky. We added image pre-processing filters and a manual dimension tweak option to keep measurements accurate.
Generating 3D Models: Converting raw pixel data into 3D shapes caused rendering glitches. We built a data converter to normalize wall coordinates before sending them to the 3D engine.
Pricing Accuracy: Sorting materials strictly by unit price wasn't enough. We adjusted our AI prompts to account for minimum purchase thresholds and local delivery fees.

What we learned

How to use computer vision to extract structured spatial data from technical 2D line drawings.
How to pair vector search with deterministic mathematical formulas so material quantities scale accurately without LLM hallucinations.
How to design an intuitive workflow specifically for non-technical contractors in the field.

What's next for CoContractor

AR On-Site Overlay: Allowing contractors to view the 3D model in Augmented Reality directly on the physical job site using mobile devices.
Direct Supplier Ordering: Adding one-click checkout so contractors can purchase ranked materials directly through local partner store APIs.
Contractor-Client Sharing Portal: Enabling contractors to send interactive 3D floor plans and material estimate quotes directly to homeowners for instant approval.