FGL-YOLO: A Lightweight and Efficient Network for Insulator Defect Detection on UAV
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Aiming at the challenges of insulator defect detection in transmission line inspection, such as complex background, small target and limited computing resources of embedded devices, this paper proposes a lightweight and efficient target detection algorithm for UAV deployment. This method makes three key improvements based on YOLOv11. Firstly, FasterNet-S0 is introduced into the backbone network, and Partial convolution (PConv) is used to reduce redundant calculations and achieve more efficient feature extraction. Secondly, Slim-neck structure and GSConv are used in the neck network to optimize the efficiency and accuracy of feature fusion. Thirdly, a lightweight Shared Convolutional Detection head (LCSD) is designed, which significantly reduces the number of parameters and enhances the recognition ability of multi-scale defects through cross-scale weight sharing, group normalization and learnable scaling module. Experiments on the public dataset IDID show that the mAP50-95 of the improved model reaches 0.5749, which is 2.47%higher than the baseline, while maintaining the real-time inference speed of 320.8 FPS. This study achieves a good balance between accuracy and efficiency, and provides a feasible solution for real-time detection of insulator defects on UAV embedded platforms.
Fields of Study & Subjects
Power Line Inspection Robots
Advanced Neural Network Applications
Infrastructure Maintenance and Monitoring
Publication details
- Journal / source
- International Conference on Communication Networks and Machine Learning (CNML)
- Pages
- 635–638
- Publication date
- 30 January 2026
- Language
- EN