Consumer LiDAR scanning is easy; converting noisy, incomplete phone scans into engineering-grade building models is not. In this talk, we share what we learned converting 150M+ square feet of iPhone/iPad scans into clean, structured geometry with semantic labels. We’ll break down the pipeline: from mesh enhancements to semantic image-based segmentation to global extraction of schematic primitives. We’ll compare classical ML segmentation with newer foundation-model approaches for semantic labeling and show where each fails on real-world imperfect data. Finally, we’ll cover a production QA strategy that utilizes human review to hit accuracy thresholds at scale. Attendees will leave with concrete techniques for building reliable scan-to-CAD/BIM systems using consumer sensors.
From iPhone Scan to Structured Building Models: Lessons from 150M Square Feet
Session Speakers
-
Anton Yakubenko
Head of AI and Computer Vision Strategy, Twindo
Anton Yakubenko, PhD, is Head of AI and Computer Vision Strategy at Twindo and an executive product leader with 20 years in 3D computer vision and spatial AI. He has founded and served as CEO of four start-ups, with two acquisitions, building products that turn reality capture (imagery and LiDAR) into structured, decision-ready data and scalable production systems with clear cost, speed and accuracy trade-offs. Anton has taken products from zero to eight-figure revenue, improved gross margins through AI plus human-in-the-loop operations and helped raise about $20M. He is an inventor on 12+ patents and author of 14 peer-reviewed papers, including an ECCV publication. Anton is a seasoned public speaker who has presented to audiences ranging from small technical workshops to 2K-person conferences.
