A Real-time LPR Deployment Scheme on Edge Devices via INT8 Quantization and TensorRT Optimization
DOI:
https://doi.org/10.65455/nygma907Keywords:
License Plate Recognition (LPR), YOLOV5, Edge Computing, INT8 QuantizationAbstract
Real-time license plate recognition (LPR) on edge devices is heavily constrained by hardware limits. This study focuses on the efficient deployment of a real-time Chinese license plate recognition system, which plays a crucial role in intelligent traffic management and automated toll collection. Despite substantial progress in artificial intelligence and deep learning algorithms, achieving real-time performance remains a challenge because of the limited computational resources of practical hardware platforms. To overcome this limitation, a two-stage model optimization approach is proposed and integrated with NVIDIA deployment toolkits to accelerate inference with only marginal accuracy loss. The proposed work delivers a high-performance license plate recognition system and demonstrates that hardware-aware model optimization is essential for achieving efficient and practical real-time deployment.
References
[1]LAROCA R, SEVERO E, ZANLORENSI L A, et al. A robust real-time automatic license plate recognition based on the YOLO detector//International Joint Conference on Neural Networks. Rio de Janeiro, Brazil: IEEE, 2018: 1-10. DOI: https://doi.org/10.1109/IJCNN.2018.8489629
[2]MASOOD S Z, SHU G, DEHGHAN A, et al. License plate detection and recognition using deeply learned convolutional neural networks [Z/OL]. arXiv:1703.07330, 2017. https://arxiv.org/abs/1703.07330.
[3]LI H, WANG P, SHEN C. Towards end-to-end car license plate detection and recognition with deep neural networks. IEEE Transactions on Intelligent Transportation Systems, 2019, 20(3): 1126-1136. DOI: https://doi.org/10.1109/TITS.2018.2847291
[4]SELMI Z, HALIMA M B, PAL U, et al. DELP-DAR system for license plate detection and recognition. Pattern Recognition Letters, 2020, 129: 213-223. DOI: https://doi.org/10.1016/j.patrec.2019.11.007
[5]MINH H T, MAI L, MINH T V. Performance evaluation of deep learning models on embedded platform for edge AI-based real-time applications//8th NAFOSTED Conference on Information and Computer Science. Hanoi, Vietnam: IEEE, 2021: 350-355.
[6]BUI T Q V, NGUYEN V T, KIM S H. Real-time object detection for autonomous driving using deep learning on edge devices. Sensors, 2021, 21(7): 2520.
[7]NAGEL M, FOURNARAKIS M, AMJAD R A, et al. A white paper on neural network quantization [Z/OL]. arXiv:2106.08295, 2021. https://arxiv.org/abs/2106.08295.
[8]NVIDIA Corporation. NVIDIA TensorRT Developer Guide. Santa Clara, CA, USA: NVIDIA Corporation, 2023.
[9]REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection//IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA: IEEE, 2016: 779-788. DOI: https://doi.org/10.1109/CVPR.2016.91
[10]REDMON J, FARHADI A. YOLO9000: Better, faster, stronger//IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 7263-7271. DOI: https://doi.org/10.1109/CVPR.2017.690
[11]XIAO B, GUO J, HE Z. Real-time object detection algorithm of autonomous vehicles based on improved YOLOv5s. Scientific Programming, 2021, 2021: 1-9. DOI: https://doi.org/10.1109/CVCI54083.2021.9661149
[12]JOCHER G, CHAURASIA A, STOKEN A, et al. ultralytics/yolov5: v7.0—YOLOv5 SOTA realtime instance segmentation [Z/OL]. Zenodo, 2022. https://doi.org/10.5281/zenodo.7347926.
[13]LI H, SHEN C. Reading car license plates using deep convolutional neural networks and LSTMs [Z/OL]. arXiv:1601.05610, 2016. https://arxiv.org/abs/1601.05610.
[14]ZHERZDEV S, GRUZDEV A. LPRNet: License plate recognition via deep neural networks [Z/OL]. arXiv:1806.10447, 2018. https://arxiv.org/abs/1806.10447.
[15]XU Z, YANG W, MENG A, et al. Towards end-to-end license plate detection and recognition: A large dataset and baseline//European Conference on Computer Vision Workshops. Munich, Germany: Springer, 2018: 261-277. DOI: https://doi.org/10.1007/978-3-030-01261-8_16
[16]KAMILARIS A, BRINK C V D, KARATSIOLIS S. Training deep learning models via synthetic data: Application in unmanned aerial vehicles//IEEE International Conference on Big Data. Los Angeles, CA, USA: IEEE, 2019: 6089-6091. DOI: https://doi.org/10.1007/978-3-030-29930-9_8
[17]GHOLAMI A, KIM S, DONG Z, et al. A survey of quantization methods for efficient neural network inference [Z/OL]. arXiv:2103.13630, 2021. https://arxiv.org/abs/2103.13630.
[18]LIANG T, GLOSSNER J, WANG L, et al. Pruning and quantization for deep neural network acceleration: A survey. Neurocomputing, 2021, 461: 370-403. DOI: https://doi.org/10.1016/j.neucom.2021.07.045
[19]ZHOU Y, YANG K. Exploring TensorRT to improve real-time inference for deep learning//IEEE International Conference on Artificial Intelligence and Computer Applications. Dalian, China: IEEE, 2022: 201-206.
[20]AGUILAR A H, BONILLA-ROBLES J C, ZAVALA-DÍAZ J C, et al. Real-time video image processing through GPUs and CUDA and its future implementation in real problems in a smart city//IEEE International Conference on Engineering Veracruz. Boca del Río, Mexico: IEEE, 2019: 1-6.
[21]GUO H, TIAN B, YANG Z, et al. DeepStream: Bandwidth efficient multi-camera video streaming for deep learning analytics//IEEE International Conference on Multimedia and Expo Workshops. Brisbane, Australia: IEEE, 2023: 1-6.
Downloads
Published
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
Issue
Section
License
Copyright (c) 2026 The Author(s). Applied Artificial Intelligence Research published by CSTDP
This work is licensed under a Creative Commons Attribution 4.0 International License.
