A lightweight single stage small object detection network based on knowledge distillation and novel feature networks

Miao Zhang; Xuefang Zhou; Junchao Hu

doi:10.1007/s11801-026-4272-4

Optoelectronics Letters ›› 2026, Vol. 22 ›› Issue (5) :282 -288. DOI: 10.1007/s11801-026-4272-4

Article

research-article

A lightweight single stage small object detection network based on knowledge distillation and novel feature networks

Author information +

History +

PDF

Abstract

With the increasing demand for traffic sign detection, the challenge of small target detection has become particularly prominent. The present study proposes an innovative approach by integrating knowledge distillation, L2 loss function, and convolutional block attention module (CBAM) mechanism to effectively tackle this issue. This series of improvements not only provide a new idea for small target detection, but also bring significant performance improvement in actual traffic scenes. Then, the integration method of the bidirectional feature pyramid network (BiFPN) is used to enhance the flexibility of the neural network to deal with input of different scales, while speeding up and improving the process of feature fusion. The experimental results demonstrate that when processing the Chinese city traffic sign detection benchmark (CCTSDB) dataset and executing the FLOW-IMG small target detection task, the optimized algorithm shows obvious performance improvement, and its accurate recognition rate jumps to 97% and 84.9%, respectively. For the basic algorithm, two datasets achieved improved accuracy by an innovative approach, improving accuracy by 5.8% and 1.3%, respectively. In terms of resource efficiency, compared to the original teacher model, the newly constructed model reduced the number of computing participants by approximately 15% during execution, while successfully reducing the overall computing task load by 14%.

Keywords

Cite this article

Download citation ▾

Miao Zhang, Xuefang Zhou, Junchao Hu. A lightweight single stage small object detection network based on knowledge distillation and novel feature networks. Optoelectronics Letters, 2026, 22(5): 282-288 DOI:10.1007/s11801-026-4272-4

登录浏览全文

4963

注册一个新账户忘记密码

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Wu Y, Zhang T, Niu J, et al.. YOLO-based lightweight traffic sign detection algorithm and mobile deployment. Optoelectronics letters, 2025, 21(4): 249-256 J]

[2]	Zaklouta F, Stanciulescu B. Real-time traffic sign recognition in three stages. Robotics & autonomous systems, 2014, 62(1): 16-24 J]

[3]	KUMAR P P, KISHEN R C, RAVIKUMAR M. Traffic sign detection and recognition using deep learning[J]. Cognitive science and technology, 2022: 345–354.

[4]	CAI Z, VASCONCELOS N. Cascade R-CNN: delving into high quality object detection[EB/OL]. (2017-12-03) [2025-07-24]. https://arxiv.org/abs/1712.00726.

[5]	RAHIMA K, MUHAMMAD H. What is YOLOv5: a deep look into the internal features of the popular object detector[EB/OL]. (2024-07-30) [2025-07-24]. https://arxiv.org/abs/2407.20892.

[6]	Sun X, Liu K, Chen L, et al.. LLTH-YOLOv5: a real-time traffic sign detection algorithm for low-light scenes. Automotive innovation, 2024, 7(1): 121-137 J]

[7]	Ding T, Feng K, Li T, et al.. An improved anchor-free detection method for traffic sign detection. 2nd International Symposium on Computer Engineering and Intelligent Communications, August 6–8, 2021, Nanjing, China, 2021348-351[C]

[8]	Guilherme M, Etienne E, Antoine G. Impacts of the CBAM on EU trade partners: consequences for developing countries. Economics, political science, 2023, 24: 243-259[J]

[9]	Zhou P, Aysa A, Ubul K. Research on knowledge distillation algorithm based on YOLOv5 attention mechanism. Expert systems with applications, 2024, 240: 122553 J]

[10]	Tian Y, Wang G, Deng X, et al.. Intelligent reconstruction algorithm of hydrogen-fueled scramjet combustor flow based on knowledge distillation model compression. International journal of hydrogen energy, 2024, 49(8): 1278-1291 J]

[11]	Abdallah S M. Real-time vehicles detection using a Tello Drone with YOLOv5 algorithm. Cihan University-Erbil scientific journal, 2024, 8(1): 1-7[J]

[12]	Hu J, Shen L, Sun G. Squeeze-and-excitation networks. Proceedings of the 2018 IEEE Conference on Computer Vision and Pattern Recognition, June 18–23, 2018, Salt Lake City, Piscataway, USA, 2018New YorkIEEE7132-7141[C]

[13]	Tan M, Pang R, Le Q V. EfficientDET: scalable and efficient object detection. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, June 13–19, 2020, Seattle, WA, USA, 2020New YorkIEEE10781-10790[C]

[14]	MÜLLER R, KORNBLITH S, HINTON G E. When does label smoothing help?[EB/OL]. (2019-06-16) [2025-07-24]. https://arxiv.org/abs/1906.02629.

[15]	Zhang J, Zou X, Kuang L, et al.. CCTSDB 2021: a more comprehensive traffic sign detection benchmark. Human-centric computing and information sciences, 2022, 12: 23[J]

[16]	Zhang J, Zheng Z, Xie X, et al.. ReYOLO: a traffic sign detector based on network reparameterization and features adaptive weighting. Journal of ambient intelligence and smart environments, 2022, 14(4): 317-334 J]

[17]	ZHAO Y, PHILIPP K. Real-time online video detection with temporal smoothing transformers[EB/OL]. (2022-09-19) [2025-07-24]. https://arxiv.org/abs/2209.09236.

[18]	Zhang J, Wang W, Lu C, et al.. Lightweight deep network for traffic sign classification. Annals of telecommunications, 2020, 75(7–8): 369-379 J]

[19]	Yang C, Dong G, Hao Y, et al.. Enhanced YOLO network for improving the efficiency of traffic sign detection. Applied sciences, 2024, 14(2): 555 J]

[20]	Cheng Y, Zhu J, Jiang M, et al.. FLOW: a dataset and benchmark for floating waste detection in inland waters. IEEE/CVF International Conference on Computer Vision, October 10–17, 2021, Montreal, QC, Canada, 2021New YorkIEEE[C]