Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 26, 2026˜The œinternational archives of the photogrammetry, remote sensing and spatial information sciences/International archives of the photogrammetry, remote sensing and spatial information sciencesOpen Access

Foundation Model-Based Pipeline for 3D Damage Localization in Built Infrastructure

View Full Paper
Ask AI
Bookmark
Share

Authors

ZYZhiya YangRLRoberto de LimaMVMaarten Vergauwen

Discussion

Loading...

Member takes

Overview

Randomized trial demonstrates data-efficient damage localization in built infrastructure, indicating improved accuracy through innovative methods.

Key Points

  • This research aims to develop a foundation model-based pipeline for cost-effective and accurate damage localization in various infrastructure types.
  • Used DINOv3 features for image-level classification.
  • Applied Grad-CAM for weak localization and SAM for prompt-guided pixel segmentation.
  • Evaluated on Sewer-ML and a custom historic masonry dataset.
  • DINOv3 classifier achieved an F2-score of 0.72 compared to 0.64 for Google ViT.
  • Heatmap-guided prompting strategy yielded a mean Dice score of 0.69 and mean IoU of 0.53.
  • Classification stage reached an F2-score of 0.99 on the masonry dataset.

Cite This Study

Yang et al. (2026) studied this question.

synapsesocial.com/papers/6a65a660d3aea3239cd77d09https://doi.org/10.5194/isprs-archives-xlix-b2-2026-1275-2026
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Maintenance-oriented tunnel digital model generation via panoptic segmentation of ultra-high-resolution images2026 · 1 citations
  2. 2From Survey to Action: AI-Driven Severe Damage Mapping2026
  3. 3Integrated UAV Path Planning and Attention-Enhanced Instance Segmentation for Automated Infrastructure Surface Defect Detection2026
  4. 4Vector-Guided Post-Earthquake Damaged Road Extraction Using Diffusion-Augmented Remote Sensing Imagery2026 · 1 citations
  5. 5Holistic image‐based analysis of damage on concrete surfaces—A multifaceted approach based on supervised machine learning2024 · 4 citations