Automated Multi-Class Structural Damage Detection and Segmentation for Unmanned Aerial Vehicle-Based Infrastructure Inspection Using Deep Learning

The research, by presenting a three-stage deep learning framework that provides fast and accurate damage classification, pixel-level localization, and real-time reporting of infrastructures such as bridges and offshore platforms using UAVs, shows how to overcome challenges like class imbalance and low-resolution images.