Making Blueprint Automation Easier for PLANX
Client Background
PLANX is a platform that helps make 3D models of rooms and buildings. It uses visualization tools and virtual reality (VR) to make designing and planning both homes and business spaces easier. PLANX allows users to see and change their designs in real time, turning ideas into detailed digital layouts.
Project Goals and Challenges
PLAR had trouble turning physical blueprints into digital versions. Manually converting these blueprints was slow and had many errors. There were two main problems:
- Scaling Problems: There were more and more blueprints, and converting them by hand was becoming impossible.
- Inconsistent Quality: The blueprints were often in bad shape. Some were blurry scans, and others were messy sketches.
The goal of the project was to create an automated solution for generating 2D plans from blueprints, which could then be used to create 3D room models.
Client
Planx
Timeline
6 months
Services
AI Development, Computer Vision, Blueprint Digitization, Image Enhancement, Object Detection, Room Classification, OCR Processing
Technological Solutions
Our team tackled these problems using artificial intelligence (AI) and computer vision. Here are the main parts of the solution.
Core Technology:
We used Python with OpenCV and Tesseract OCR to process images and text. We also made a custom U-NET model to extract information from the blueprints and turn them into vector graphics.
Primary avancements:
- Improving Image Quality: OpenCV fixed common problems with blueprints, like crooked lines and shadows.
- Line Detection and Vectorization: The U-NET model found and traced important parts of the blueprint like walls, windows, and doors.
- Object Detection: The system could find and label things like furniture to give more details about the room.
- Optimized Response Times: Fast processing through a microservice framework.
- Room Classification: The system could identify different room types based on the layout and objects inside.
- Text Extraction: The OCR feature read the text on the blueprints, like dimensions and notes.
- Output Formatting: The final output was turned into standard formats like SVG that could be used with different 3D modeling tools.
Results
- Faster Conversion: The time to convert blueprints went from days to just a few hours.
- Better Scalability: Automation made it possible to process hundreds of blueprints every day.
- Higher Accuracy: Custom AI models helped reduce errors, making the digital models much better.