Work

Automated pavement condition rating for cycle routes and greenways using deep learning

Year: 2026

Type: article

Abstract: Active travel routes promote sustainability, public health, and safety while reducing environmental impact. Efficient maintenance of cycling infrastructure requires proactive assessment of pavement conditions. This paper presents an intelligent pavement-rating framework that uses deep learning to automatically evaluate offline cycleways using GoPro footage captured from a handlebar-mounted camera. A new Cycle Route Surface Index (CRSI) rating scale was developed specifically for cycleways and greenways in collaboration with Transport Infrastructure Ireland (TII). Using this scale, TII experts manually annotated 43,594 greenway images to create a benchmark dataset. Deep learning models were then developed and optimized to classify pavement conditions, achieving substantially higher accuracy and computational efficiency than traditional approaches. The best-performing model achieved 99.9% test accuracy on the CRSI dataset. To enhance transparency and trustworthiness, the framework incorporates transformer-based Grad-CAM to provide visual explanations of model predictions, supporting maintenance teams and planners in making timely, informed decisions. The proposed framework enables proactive greenway management and demonstrates strong potential for future deployment.

Source: Automation in Construction

DOI: 10.1016/j.autcon.2026.107210

Authors: Syed M.H. Shah, Waqar S. Qureshi, Gerard O’ Dea, David Power, Ihsan Ullah

Institutions: Ollscoil na Gaillimhe – University of Galway, Transport Infrastructure Ireland

Language: English

Topic: Infrastructure Maintenance and Monitoring

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Industry, innovation and infrastructure, Sustainable cities and communities

Open Access status: hybrid

Estimated APC paid: $5,490