Abnormal behavior recognition of experimental process based on improved YOLOV4

Changyong Zhang; Chunyang Jin; Yuzhou Li

doi:10.1117/12.2682353

26 May 2023 Abnormal behavior recognition of experimental process based on improved YOLOV4

Changyong Zhang, Chunyang Jin, Yuzhou Li

Proceedings Volume 12700, International Conference on Electronic Information Engineering and Data Processing (EIEDP 2023); 1270023 (2023) https://doi.org/10.1117/12.2682353
Event: International Conference on Electronic Information Engineering and Data Processing (EIEDP 2023), 2023, Nanchang, China

Abstract

In order to strengthen the management of experimental teaching process and improve the accuracy of abnormal behavior detection of students, the existing YOLOv4 algorithm is improved. By adding channel attention mechanism SE module in the main feature extraction network, the algorithm ensures fast and effective extraction of student behavior effective features. A soft pooled SPP network based on CSP structure is used to assist optimization and feature extraction to reduce information loss in the pooling process. By introducing Focal loss function to deal with uneven number of plus and minus samples and difficult sample training in data, the effect of student behavior classification is improved. The algorithm is verified by experiment on the self-made behavior data set. The results show that the improved YOLOv4 algorithm has a better detection effect on students' behavior, and the average accuracy rate reaches 86.41%. Compared with the YOLOv4 algorithm, the recognition accuracy rate is improved by about 7%.

Citation Download Citation

Changyong Zhang, Chunyang Jin, and Yuzhou Li "Abnormal behavior recognition of experimental process based on improved YOLOV4", Proc. SPIE 12700, International Conference on Electronic Information Engineering and Data Processing (EIEDP 2023), 1270023 (26 May 2023); https://doi.org/10.1117/12.2682353

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