Case Study

Blueprint AI — Computer Vision & OCR for Architectural Blueprints

A computer vision and OCR pipeline that detects countertops and appliances in architectural blueprints, reads the drawing scale, and turns pixels into real-world dimensions and cost estimates.

Focus: object detection, instance segmentation, OCR scale extraction, and measurement-driven cost estimation.

Problem

Estimating countertop material from a blueprint is slow, manual work: someone reads the scale, measures each surface, and builds a cost sheet by hand. The goal was to automate that from the drawing itself.

Detection

Trained Faster R-CNN object detection on 10,000+ annotated blueprint images to locate countertops, sinks, and appliances with bounding boxes and confidence scores.

Segmentation

Trained Mask R-CNN (ResNet-50-FPN) for instance segmentation, producing pixel-accurate masks that isolate the exact boundary of each countertop rather than a rectangle around it.

Scale extraction and measurement

Used docTR OCR to read the blueprint scale text, computed inches-per-pixel at 200 DPI, and converted segmented regions into real-world dimensions, area estimates, and a per-object material cost table.

Outcome

A working end-to-end pipeline with a Streamlit demo front end: upload a blueprint, get detected objects, measured surfaces, and an auto-generated cost estimate.

Screenshots

Blueprint detection: countertop and appliance bounding boxes with confidence scores

Instance segmentation: Mask R-CNN masks showing exact countertop boundaries

OCR scale extraction computing inches per pixel and a per-object cost table

Full floor plan analysis with scale extraction and cost estimation

Core technologies

Python, PyTorch, Faster R-CNN, Mask R-CNN, OpenCV, docTR OCR, Streamlit.

More case studies