Optimized YOLOv11n for real-time behavior detection in classrooms
Jiangteng Long
Faculty of Information Technology, City University Malaysia/Jiangxi Vocational & Technical College of Information Application
M. Kazem Chamran
Faculty of Information Technology, City University Malaysia
Jiaojie Song
Jiangxi Vocational & Technical College of Information Application
DOI: https://doi.org/10.59429/esta.v13i3.15550
Keywords: image recognition; YOLOv11n; real-time response; deep learning; human pose estimation
Abstract
This research proposes a multi-strategy joint optimization framework for classroom behavior detection based on the improved YOLOv11n. It combines the Normalized Wasserstein Distance (NWD), Shape - Intersection over Union (Shape-IoU), and Shape-NWD fusion loss function, and introduces the C2PSA attention enhancement module and a full - lifecycle privacy protection mechanism. In response to the problems of gradient vanishing and inaccurate matching of the traditional CIoU loss for small targets, this paper introduces the NWD index, replacing the overlap - based similarity measure with the Gaussian distribution distance to effectively stabilize the training gradient of sparse - pixel small targets.
References
[1] O. Zawacki-Richter, V.I. Marín, M. Bond and F. Gouverneur, "Systematic Review of Research on Artificial Intelligence Applications in Higher Education–Where Are the Educators?" Int. J. Educ. Technol. High. Educ., vol. 16, no. 1, pp. 1-27, 2019, doi: 10.1186/s41239-019-0171-0.
[2] R. Khanam and M. Hussain, "YOLOv11: An Overview of the Key Architectural Enhancements," Corr, vol. abs/2410.17725, pp. 1-9, 2024.
[3] F. Yang, "SCB-dataset: A Dataset for Detecting Student Classroom Behavior," Arxiv (Cornell University), vol. abs/2304.02488, 2023, doi: 10.48550/arxiv.2304.02488.
[4] Y. Zhou, K. Cao, De Li and J. Piao, "Fine-YOLO: A Simplified X-ray Prohibited Object Detection Network Based on Feature Aggregation and Normalized Wasserstein Distance," Sensors, vol. 24, 2024, doi: 10.3390/s24113588.
[5] H. Zhang and S. Zhang, "Shape-IoU: More Accurate Metric Considering Bounding Box Shape and Scale.," Corr, vol. abs/2312.17663, 2023, doi: 10.48550/arxiv.2312.17663.
[6] C. Rao, W. Li, X. Xie and G. Cheng, "A Shape-Based Quadrangle Detector for Aerial Images.," Pattern Recognition and Computer Vision, Prcv 2023, Pt IV, vol. 14428, pp. 368-379, 2024, doi: 10.1007/978-981-99-8462-6_30.
[7] F. Yang and T. Wang, "SCB-Dataset3: A Benchmark for Detecting Student Classroom Behavior.," Corr, vol. abs/2310.02522, 2023, doi: 10.48550/arxiv.2310.02522.
[8] I. Bartoletti, "AI in Education," The Ethics of Artificial Intelligence in Education, pp. 74-90, 2022, doi: 10.4324/9780429329067-5.