Lightweight Detection of Dangerous Driving Features via Knowledge Distillation
DOI:
https://doi.org/10.61173/j9ee3k30Keywords:
Dangerous driving detection, Knowledge distillation, Lightweight model, EfficientNet-V2, Mo-bileNet-V3Abstract
To address the contradiction between accuracy and efficiency in dangerous driving detection models, this study proposes a lightweight feature learning framework based on knowledge distillation. By constructing a collaborative model architecture, with EfficientNet-V2 as the knowledge source and MobileNet-V3 as the lightweight carrier, combined with an attention feature transfer strategy, efficient extraction of key features of driving behaviors is achieved. Experiments based on the Kaggle Driver Inattention Detection Dataset verify that this method reduces computational demand while increasing operating speed and accuracy. The research results can provide low-latency, high-robustness behavior monitoring solutions for in-vehicle embedded systems.
References
[1] Hinton, G. E., Vinyals, O., & Dean, J. (2015). Distilling the knowledge in a neural network. arXiv: Machine Learning. https://arxiv.org/abs/1503.02531
[2] Garcıa-Alcaide, D. C., et al. (n.d.). Analyzing model behavior for driver emotion recognition and drowsiness detection using explainable artificial intelligence.
[3] Rehman, A., Waseem, M. Z., Rafey, A., Hussaini, A. A., Parveen, H., & Khan, N. (2025). Identification detection and YoloV5 based driver drowsiness framework. International Journal of Innovative Science and Research Technology, 10, 2806-2815.
[4] Zhao, C., Gao, Z., Wang, Q., Xiao, K., Mo, Z., & Deen, M. J. (2023). FedSup: A communication-efficient federated learning fatigue driving behaviors supervision approach. Future Generation Computer Systems, 138, 52-60. https://doi. org/10.1016/j.future.2022.08.009
[5] Liu, J., Li, T., Yuan, Z., Huang, W., Xie, P., & Huang, Q. (2022). Symbolic aggregate approximation based data fusion model for dangerous driving behavior detection. Information Sciences, 609, 626-643. https://doi.org/10.1016/ j.ins.2022.07.118
[6] Liu, J., Huang, W., Li, H., Ji, S., Du, Y., & Li, T. (2023). SLAFusion: Attention fusion based on SAX and LSTM for dangerous driving behavior detection. Information Sciences, 640, 119063. https://doi.org/10.1016/j.ins.2023.119063
[7] Song, W., Zhang, G., & Long, Y. (2023). Identification of dangerous driving state based on lightweight deep learning model. Computers and Electrical Engineering, 105, 108509. https://doi.org/10.1016/j.compeleceng.2022.108509
[8] Shahverdy, M., Fathy, M., Berangi, R., & Sabokrou, M. (2020). Driver behavior detection and classification using deep convolutional neural networks. Expert Systems with Applications, 149, 113240. https://doi.org/10.1016/ j.eswa.2020.113240
[9] Hou, M., Wang, M., Zhao, W., Ni, Q., Cai, Z., & Kong, X. (2022). A lightweight framework for abnormal driving behavior detection. Computer Communications, 184, 128-136. https://doi. org/10.1016/j.comcom.2021.12.007
[10] Pan, H., Guan, S., & Zhao, X. (2024). LVD-YOLO: An efficient lightweight vehicle detection model for intelligent transportation systems. Image and Vision Computing, 151, 105276. https://doi.org/10.1016/j.imavis.2024.105276
[11] Ni, W., & Bai, L. (2025). Improved coordinate attention network for classification of dangerous driving behavior. Franklin Open, 10, 100219. https://doi.org/10.1016/ j.fraope.2025.100219
[12] Khan, S., Siddique, T. H. M., Ibrahim, M. S., Siddiqui, A. J., & Huang, K. (2025). Spatio-temporal deep learning for improved face presentation attack detection. Knowledge-Based Systems, 311, 113059. https://doi.org/10.1016/j.knosys.2025.113059
[13] Feng, C., Zhu, S., Tang, M., Zhao, H., Yuan, Q., & Wang, B. (2025). Study on image acquisition and camera positioning of depth recognition model in the tobacco curing stage. Engineering Applications of Artificial [1]Intelligence, 143, 109992. https:// Dean&Francis Yiming Yang doi.org/10.1016/j.engappai.2024.109992
[14] Zhao, L., Yang, F., Bu, L., Han, S., Zhang, G., & Luo, Y. (2021). Driver behavior detection via adaptive spatial attention mechanism. Advanced Engineering Informatics, 48, 101280. https://doi.org/10.1016/j.aei.2021.101280
[15] Feng, Z., Wei, X., Bi, Y., Zhu, D., & Huang, Z. (2025). An integrated framework for driving risk evaluation that combines lane-changing detection and an attention-based prediction model. Traffic Injury Prevention, 26(2), 198-206. https://doi.org/ 10.1080/15389588.2024.2399301
[16] Chen, Z., Fu, L., Yao, J., Guo, W., Plant, C., & Wang, S. (2023). Learnable graph convolutional network and feature fusion for multi-view learning. Information Fusion, 95, 109-119. https://doi.org/10.1016/j.inffus.2023.02.013
[17] Liu, J., Yang, N., Lee, Y., Huang, W., Du, Y., Li, T., & Zhang, P. (2024). FedDAF: Federated deep attention fusion for dangerous driving behavior detection. Information Fusion, 112, 102584. https://doi.org/10.1016/j.inffus.2024.102584
[18] Zhang, J., Zhang, Y., Shao, F., Ma, X., Feng, S., Wu, Y., & Zhou, D. (2024). Efficient face anti-spoofing via head-aware transformer based knowledge distillation with 5 MB model parameters. Applied Soft Computing, 166, 112237. https://doi. org/10.1016/j.asoc.2024.112237
[19] Yaremchenko, O., Pukach, P., & Karovic, V. (2025). Enhancing micro-expression detection methods based on machine learning. Procedia Computer Science, 257, 769-776. https://doi.org/10.1016/j.procs.2025.03.099
[20] DeVoe, K., Takahashi, G., Tarshizi, E., & Sacker, A. (2024). Evaluation of the precision and accuracy in the classification of breast histopathology images using the MobileNetV3 model. Journal of Pathology Informatics, 15, 100377. https://doi. org/10.1016/j.jpi.2024.100377
[21] Rahman, A., Hriday, M. B. H., & Khan, R. (2022). Computer vision-based approach to detect fatigue driving and face mask for edge computing device. Heliyon, 8(10), e11204. https://doi.org/10.1016/j.heliyon.2022.e11204
[22] Zhao, Z., Bakar, E. B. A., Razak, N. B. A., & Akhtar, M. N. (2024). Corrosion image classification method based on EfficientNetV2. Heliyon, 10(17), e36754. https://doi. org/10.1016/j.heliyon.2024.e36754
[23] Ni, W., & Bai, L. (2025). Improved coordinate attention network for classification of dangerous driving behavior. Franklin Open, 10, 100219. https://doi.org/10.1016/ j.fraope.2025.100219
[24] Zhao, S., et al. (2025). Lightweight YOLOM-Net for automatic identification and real-time detection of fatigue driving. Computers, Materials & Continua, 82(3), 4995-5017. https://doi.org/10.32604/cmc.2025.059972
[25] Tomar, A. S., Arya, K. V., & Rajput, S. S. (2025). Learning face super-resolution through identity features and distilling facial prior knowledge. Expert Systems with Applications, 262, 125625. https://doi.org/10.1016/j.eswa.2024.125625
[26] Tran, H. N., Nguyen, N. N. N., Le, N. Q. P., Le, T. A. N., & Nguyen, A. D. (2025). Grounding DINO and distillationenhanced model for advanced traffic sign detection and classification in autonomous vehicles. Engineering Science and Technology, an International Journal, 64, 102028. https://doi. org/10.1016/j.jestch.2025.102028
[27] Chai, X., Zhao, M., Li, J., & Li, J. (2025). Image small target detection in complex traffic scenes based on Yolov8 multiscale feature fusion. Alexandria Engineering Journal, 126, 578-590. https://doi.org/10.1016/j.aej.2025.04.105
[28] Tang, Y. M., Zhao, D., Chen, T., & Fu, X. (2025). A systematic review of abnormal behaviour detection and analysis in driving simulators. Transportation Research Part F: Traffic Psychology and Behaviour, 109, 897-920. https://doi. org/10.1016/j.trf.2025.01.002
Downloads
Published
Issue
Section
License
Copyright (c) 2025 by the authors.
This work is licensed under a Creative Commons Attribution 4.0 International License.
