Computer vision for construction blueprints and a rebuilt 2D takeoff platform, reducing rendering time by more than 50%.
Computer Vision
Rendering Performance
Construction
50%+
Rendering Time Reduction
Led development of a computer-vision pipeline for construction takeoffs, combining rotation-invariant template matching in OpenCV, SAM2-based segmentation of tiled CAD blueprints, and GPU-parallelized processing.
Built wall detection and symbol recognition for commercial construction blueprints, combining Hough Transform approaches with skeleton extraction for more accurate parsing.
Rebuilt the 2D takeoff platform, cutting rendering times by more than 50% and enabling large PDF blueprint projects to run on consumer hardware.