π¬ AI predicts road damage with deep learning
Researchers introduced an interpretable dual-stage Deep Learning framework to model pavement deterioration severity and physical extent. This work was conducted by several authors at various institutions, including Amirkabir University of Technology and EPFL. The system achieved F1-scores up to 0.982 and R2 values like 0.954 across different crack and pothole metrics. The Explainable AI component confirmed that high-severity alligator cracking influences pothole area, aligning with known pavement mechanisms. Further external validation is needed before deploying this framework for sparse pothole targets in LTPP planning. π§