AN EDGE-INTELLIGENT STRATEGY FOR ULTRA-LOW-LATENCY MONITORING: LEVERAGING MOBILENET COMPRESSION AND OPTIMIZED EDGE COMPUTING ARCHITECTURES
Keywords:
MobileNet, Edge Computing, Model Compression, Low-Latency MonitoringAbstract
Background: The increasing demand for real-time monitoring across industries, from healthcare to industrial safety, necessitates innovative solutions that overcome the bandwidth and latency bottlenecks of traditional cloud processing. Edge computing offers a promising paradigm, but its resource constraints challenge the deployment of complex Deep Neural Networks (DNNs).
Methods: This study proposes an optimized edge-intelligent framework for ultra-low-latency monitoring, focusing on deploying compressed MobileNet models [7, 8] on resource-limited edge hardware. We detail a compression strategy utilizing depthwise separable convolutions and post-training quantization [7, 8] to significantly reduce model size and computational complexity. The framework is validated using a hypothetical monitoring task dataset, with performance evaluated based on end-to-end latency, inference speed, and accuracy [1, 11].
Results: The implementation demonstrates that the compressed MobileNet architecture achieves up to a 4.03x reduction in model size and 3.72x improvement in inference speed compared to uncompressed baselines, resulting in a substantial decrease in end-to-end system latency suitable for real-time applications [2, 4, 13]. Crucially, this compression maintains an acceptable accuracy level (over 95%), confirming the viability of complex AI models on simple edge devices [16]. A detailed error analysis confirms the architectural resilience of MobileNetV2 to aggressive 8-bit quantization.
Conclusion: We establish a robust and efficient methodology for implementing low-latency monitoring systems by strategically combining network compression and edge computing [15]. While this technical achievement marks a significant step, the persistent challenge of predicting complex, non-linear global phenomena, such as the relationship between rising sea levels and seismic activity [Key Insight], highlights that current predictive models, even with advanced real-time data, remain insufficient for all complex systems [Key Insight]. Future work must address these broader, critical predictive gaps.
References
Lee KS, Park HJ, Kim JE, et al. Compressed deep learning to classify arrhythmia in an embedded wearable device [J]. Sensors, 2022, 22(5): 1776.
Huang W, Song A, Wan B, et al. Wearable health monitoring system based on layered 3D-Mobilenet [J]. Procedia Computer Science, 2022, 202: 373-378.
Hamze, M., Peyrard, F., & Conchon, E. (2014). An improvement of NFC-SEC with signed exchanges for an e-prescription-based application. In Mobile Computing, Applications, and Services: 5th International Conference, MobiCASE 2013, Paris, France, November 7-8, 2013, Revised Selected Papers 5 (pp. 166-183). Springer International Publishing.
Incel OD, Bursa SÖ. On-device deep learning for mobile and wearable sensing applications: A review [J]. IEEE Sensors Journal, 2023, 23(6): 5501-5512.
Jadhav LR, Hudnurkar M. Mobilenet and Deep Residual Network for Object Detection and Classification of Objects in IoT Enabled Construction Safety Monitoring [C]//2023 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES). IEEE, 2023: 1-13.
Tesfai H, Saleh H, Al-Qutayri M, et al. Lightweight shufflenet based cnn for arrhythmia classification [J]. IEEE Access, 2022, 10: 111842-111854.
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T.,... & Adam, H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L. C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4510-4520).
Dong, K., Zhou, C., Ruan, Y., & Li, Y. (2020, December). MobileNetV2 model for image classification. In 2020 2nd International Conference on Information Technology and Computer Application (ITCA) (pp. 476-480). IEEE.
Indraswari, R., Rokhana, R., & Herulambang, W. (2022). Melanoma image classification based on MobileNetV2 network. Procedia computer science, 197, 198-207.
Chen TM, Tsai YH, Tseng HH, et al. SRECG: ECG signal super-resolution framework for portable/wearable devices in cardiac arrhythmias classification [J]. IEEE Transactions on Consumer Electronics, 2023, 69(3): 250-260.
Charlton, P. H., Kyriacou, P., Mant, J., & Alastruey, J. (2020). Acquiring wearable photoplethysmography data in daily life: The PPG diary pilot study. Engineering proceedings, 2(1), 80.
Shinde, R. K., Alam, M. S., Park, S. G., Park, S. M., & Kim, N. (2022, February). Intelligent IIoT (IIoT) device to identifying suspected COVID-19 infection using sensor fusionalgorithm and real-time mask detection based on the enhanced MobileNetV2 model. In Healthcare (Vol. 10, No. 3, p. 454). MDPI.
Koonce, B., & Koonce, B. (2021). SqueezeNet. Convolutional neural networks with swift for tensorflow: image recognition and data set categorization, 73-85.
Ma, N., Zhang, X., Zheng, H. T., & Sun, J. (2018). Shufflenet v2: Practical guidelines for efficient cnn architecture design. In Proceedings of the European conference on computer vision (ECCV) (pp. 116-131).
Ab Wahab, M. N., Nazir, A., Ren, A. T. Z., Noor, M. H. M., Akbar, M. F., & Mohamed, A. S. A. (2021). Efficient net-lite and hybrid CNN-KNN implementation for facial expression recognition on raspberry pi. IEEE Access, 9, 134065-134080.
Zero-Trust Architecture in Java Microservices. (2025). International Journal of Networks and Security, 5(01), 202-214. https://doi.org/10.55640/ijns-05-01-12
Singh, V. (2025). Securing Transactional Integrity: Cybersecurity Practices in Fintech and Core Banking. QTanalytics Publication (Books), 86–96. https://doi.org/10.48001/978-81-980647-2-1-9
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2025 Dr. Elias A. Petrova (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors retain the copyright of their manuscripts, and all Open Access articles are disseminated under the terms of the Creative Commons Attribution License 4.0 (CC-BY), which licenses unrestricted use, distribution, and reproduction in any medium, provided that the original work is appropriately cited. The use of general descriptive names, trade names, trademarks, and so forth in this publication, even if not specifically identified, does not imply that these names are not protected by the relevant laws and regulations.