Skip to content
ZEYN

IAAC — Institute for Advanced Architecture of Catalonia (MRAC) · ACADEMIC · RESEARCH

AI-Driven Structural Diagnostics & Stylized Point Cloud Reconstruction

Drone Photogrammetry, YOLO Defect Detection, AI Style Transfer

Context
IAAC — Institute for Advanced Architecture of Catalonia (MRAC) — Academic Research
Year
2024
Practice
RESEARCH
Location
Barcelona, Spain
Status
COMPLETED
Point cloud of the IAAC building with defect analytics dashboard

Overview

A combined AI, human–machine interaction and point-cloud workflow for assessing building structural health and reconstructing it stylized. YOLO object detection identifies truss cracks, water stains and roof damage; detection results are mapped onto the 3D point cloud for structural-health visualisation. CycleGAN / Stable Diffusion style transfer then rebuilds the structure artistically — offering a digital workflow for architectural heritage preservation and intelligent urban design.

Dense point cloud model from drone photogrammetry
YOLO defect detection results mapped on the point cloud
AI style transfer reconstruction comparison

Responsibilities

  • Full-building image capture by drone (DJI) with planned flight paths over roof and truss zones
  • Dense point-cloud generation in Agisoft Metashape, including fisheye-correction for modelling accuracy
  • Dataset building and annotation; YOLOv8 training on Roboflow for cracks, roof damage and water stains
  • Stylized reconstruction with MidJourney / Stable Diffusion / Runway / ControlNet, re-modelled in Metashape

Methodology

  • Detection results mapped onto the 3D point cloud to visualise structural health in space
  • Open3D point-cloud analysis and geometric segmentation
  • Original vs stylized point clouds compared for geometric accuracy and expressive effect