Related Experiment Video
Updated: Jan 8, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
991
A multi-scale adaptive framework for high-precision rail track damage detection via StarNet and bidirectional feature
Yanzhi Pang1, Xiang Wang2, Yang Tang3
1Guangxi Key Laboratory of International Join for China-ASEAN Comprehensive Transportation, Nanning University, Nanning, 530200, China. pangyanzhi@unn.edu.cn.
Scientific Reports
|December 18, 2025
Summary
A new SNBF-YOLO model enhances railway track defect detection using Star Net and BiFPN modules. This improves accuracy and efficiency for safer transportation systems.
Area of Science:
- Engineering
- Computer Science
- Artificial Intelligence
Background:
- Railway track integrity is crucial for transportation safety.
- Manual inspection methods are inefficient and prone to errors.
- Existing deep learning models face challenges with small targets, complex backgrounds, and multi-scale defects.
Purpose of the Study:
- To develop an improved deep learning framework for accurate railway track damage detection.
- To enhance feature representation and multi-scale feature fusion for better defect identification.
Main Methods:
- Proposed SNBF-YOLO framework integrating Star Net and BiFPN modules.
- Star Net enhances feature representation by adaptively enlarging the receptive field.
- BiFPN optimizes bidirectional multi-scale feature fusion.
Main Results:
- SNBF-YOLO improved precision by 19.4%, recall by 13.2%, and mAP by 14.3% over YOLOv10n.
- Achieved higher robustness and computational efficiency for real-time detection.
- Successfully detected fine cracks and missing fasteners.
Conclusions:
- The SNBF-YOLO framework offers a robust and efficient solution for railway track defect detection.
- Further research should focus on dataset expansion and lightweight deployment for broader applicability.
- The model shows promise for enhancing railway safety through advanced AI.