Work

XRoadNet: An Explainable Multi-Scale YOLOv11-Attention Fusion Framework for Real-Time Road Damage Detection and Severity Assessment

Year: 2026

Type: conference-paper

Abstract: Road defect detection is a vital component of smart transport. Conventional detection systems are expensive, cumbersome, and susceptible to errors due to manual intervention. In order to address these challenges, this paper presents XRoadNet - an explainable deep learning architecture for efficient detection and classification of road defects. XRoadNet uses image enhancement, multi-scale YOLOv11 detection, and attention-based feature extraction to enhance detection accuracy of defects under different circumstances. The severity analysis tool categorizes detected road damages based on their repair importance, whereas Grad-CAM++ is used to generate visual explanations. XRoadNet model secured an mAP@50 of 97.1%, a precision of 96.8%, and a recall of 96.4%, which signifies its efficacy in detecting road damages.

Source: International Conference Computational Vision and Bio Inspired Computing

DOI: 10.1109/iccvbic71195.2026.11688394

Authors: J. Jeslin Shanthamalar, E Gomathi, Krithika V, P Swetha, B Naveenkumar, Kalaivani T

Institutions: Sathyabama Institute of Science and Technology, Karlsruhe Institute of Technology, Sri Eshwar College of Engineering

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

OpenAlex ID: W7213554146

FWCI:

Cites: 12

Cited by:

Related to: 10