A Survey on Pruning Methods and Attention Mechanisms for YOLOv8-Based Models

Zhexu Zhang

doi:10.54691/hwg3nq26

Authors

Zhexu Zhang College of Electronic Information and Communications, Beijing University of Technology, Beijing 100124, China

DOI:

https://doi.org/10.54691/hwg3nq26

Keywords:

YOLOv8; Structured Pruning; Attention Mechanism; Knowledge Distillation; C2f Module.

Abstract

The C2f module in YOLOv8 enhances multi-scale feature fusion, yet parameter redundancy still hinders its deployment on edge devices. Current model compression techniques often overlook two critical issues: (i) topological mismatch among branched network paths, and (ii) lack of synergy between pruning criteria and distillation loss functions. This survey systematically reviews recent YOLOv8 lightweight adaptation efforts, critically analyzing limitations of existing structured pruning, attention integration, and knowledge distillation methods from two perspectives: topological coupling of network architecture and coordination of multi-objective optimization. Building upon the internal mechanics of the C2f module, we formally formulate a path-level channel dependency graph and propose a joint evaluation metric that integrates channel importance with distillation sensitivity. This provides a viable technical pathway toward lossless, high-ratio compression of YOLOv8.

Downloads

Download data is not yet available.

References

[1] Cheng, Y. (2024). YOLO V8 network lightweight method based on pruning and quantization. Mathematical Modelling and Applied Analysis, 3(2), 44–51. https://doi.org/10.54097/4qsv1565

[2] Ullah, S. S., Zamir, M. Z., Ishfaq, A., & Khan, S. (2026). Attention augmented YOLOv8 with ghost convolution for real time vehicle detection in intelligent transportation systems. arXiv preprint arXiv:2604.22856.

[3] Sabaghian, M., Keyvanrad, M. A., & Moghadami, S. M. (2025). A novel compression framework for YOLOv8. arXiv preprint arXiv:2509.12918.

[4] Zhang, Y., Zhang, J., Du, F., Kang, W., Wang, C., & Li, G. (2025). Real time lightweight vehicle object detection via layer adaptive model pruning. Electronics, 14(21), 4149. https://doi.org/10.3390/electronics14214149

[5] Ruan, J., et al. (2024). SimYOLO: A lightweight steel defect detection model based on improved YOLOv8. In Proceedings of the 2024 4th International Conference on Artificial Intelligence, Big Data and Algorithms (AIBDA). ACM.

[6] Liu, Z., Li, J., Shen, Z., Huang, G., Yan, S., & Zhang, C. (2017). Learning efficient convolutional networks through network slimming. In Proceedings of the IEEE International Conference on Computer Vision (ICCV) (pp. 2736–2744).

[7] Woo, S., Park, J., Lee, J. Y., & Kweon, I. S. (2018). CBAM: Convolutional block attention module. In Proceedings of the European Conference on Computer Vision (ECCV) (Vol. 11211, pp. 3–19).

[8] Wang, Q., Wu, B., Zhu, P., Li, P., Zuo, W., & Hu, Q. (2020). ECA Net: Efficient channel attention for deep convolutional neural networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 11534–11542).

[9] Yang, L., Zhang, R. Y., Li, L., & Xie, X. (2021). SimAM: A simple, parameter free attention module for convolutional neural networks. In Proceedings of the 38th International Conference on Machine Learning (ICML) (Vol. 139, pp. 11863–11874).

[10] Hou, Q., Zhou, D., & Feng, J. (2021). Coordinate attention for efficient mobile network design. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 13713–13722).

[11] Lim, C., & Lee, S. (2024). Effectiveness of attention mechanisms in YOLOv8 for maritime vessel detection. Journal of Marine Science and Engineering, 14(5), 433. https://doi.org/10.3390/jmse14050433

[12] Zhang, D., Wang, Y., Li, S., Li, S., & Wang, Y. (2026). Knowledge distillation in object detection: A survey from CNN to transformer. Sensors, 26(1), 292. https://doi.org/10.3390/s26010292

[13] Huang, J., Ma, Z., Wu, Y., Bao, Y., Wang, Y., Su, Z., & Guo, L. (2025). YOLOv8 DDS: A lightweight model based on pruning and distillation for early detection of root mold in barley seedling. Information Processing in Agriculture. https://doi.org/10.1016/j.inpa.2025.07.004

[14] Fang, G., Ma, X., Song, M., Mi, M. B., & Wang, X. (2023). DepGraph: Towards any structural pruning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 16091–16101).

[15] Kang, M. (2023). BGF YOLO: Enhanced YOLOv8 with multiscale attentional feature fusion for brain tumor detection. arXiv preprint arXiv:2309.12585.