How Much Can a 2-Pixel Error Distort a Floor-Plan Estimate?
Automated quantity takeoff finds rooms in a floor plan and measures them. I built quantity-lab to separate where the errors come from: the measuring code, a controlled geometry error, and a real room-finding step. On three held-out CubiCasa5K plans, a 2-pixel boundary error shifted room areas by 3.3 to 4.1%, a shifted plan kept its areas, and one missing divider left the total area almost unchanged while the room count went wrong. A classical OpenCV room finder matched almost none of the rooms, yet on one plan its total area was within 1.3%.
computer-visionconstructiongeometryevaluation
10 min read