Insulator detection based on improved YOLOv5s

Tong Wang; Yuan Sun

doi:10.1117/12.2685681

15 August 2023 Insulator detection based on improved YOLOv5s

Tong Wang, Yuan Sun

Proceedings Volume 12719, Second International Conference on Electronic Information Technology (EIT 2023); 127193K (2023) https://doi.org/10.1117/12.2685681
Event: Second International Conference on Electronic Information Technology (EIT 2023), 2023, Wuhan, China

Abstract

The YOLO series is a single-stage detection model, and the insulator is in a complex environment, which makes the detection accuracy generally low. At the same time, due to the large number of insulator image data, the training time of data is too long, resulting in low model training efficiency, slow detection speed and other problems. Based on YOLOv5s, this paper proposes an improved YOLOv5s insulator detection algorithm, YOLOv5c, to achieve high precision and real-time rapid detection. The classical YOLOv5s model is selected as the basis of this algorithm. Add CBAM attention mechanism to backbone network so that important features can be extracted adaptively to improve feature detection ability and detection accuracy. Add a weighted bidirectional feature pyramid network (BiFPN) to the Neck network for bidirectional cross scale connectivity and weighted feature fusion, so as to improve the efficiency of model training. The experimental results show that compared with the original algorithm, YOLOv5c has an average detection accuracy improvement of 1%, reaching 99.5%, and the training efficiency of the model is also improved by about double compared with the original model.

Citation Download Citation

Tong Wang and Yuan Sun "Insulator detection based on improved YOLOv5s", Proc. SPIE 12719, Second International Conference on Electronic Information Technology (EIT 2023), 127193K (15 August 2023); https://doi.org/10.1117/12.2685681

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available