{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,11]],"date-time":"2026-08-11T05:28:08Z","timestamp":1786426088391,"version":"build-2736575974"},"reference-count":30,"publisher":"MDPI AG","issue":"16","license":[{"start":{"date-parts":[[2026,8,8]],"date-time":"2026-08-08T00:00:00Z","timestamp":1786147200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100015688","name":"\u00d3buda University","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100015688","id-type":"DOI","asserted-by":"crossref"}]},{"id":[{"id":"https:\/\/ror.org\/00ax71d21","id-type":"ROR","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Sensors"],"abstract":"<jats:p>The deployment of deep learning-based vision systems on edge devices introduces a complex trade-off between computational efficiency and detection accuracy. In this work, we investigate this trade-off in the context of a multi-stage Automatic License Plate Recognition (ALPR) pipeline, evaluated in two heterogeneous edge execution environments: a general-purpose Raspberry Pi 5 single-board computer and the ARTPEC-8 system-on-chip integrated into an Axis smart camera, where neural network inference is accelerated by the on-chip DLPU. All experiments were performed using pre-recorded images loaded from the file system; neither the Axis camera sensor nor a live video stream was used. This study evaluates the impact of model architecture, numerical precision, and input resolution on both inference latency and detection performance. YOLOv5- and YOLOv8-based models were analyzed under multiple quantization schemes (FP32, FP16, dynamic, and INT8), while a cross-platform benchmark was conducted to assess the benefits and limitations of hardware acceleration. The results show that dedicated accelerators provide significant latency reduction at higher resolutions; however, this advantage is accompanied by reduced flexibility and increased sensitivity to quantization effects. In contrast, CPU-based execution enables the use of more recent and quantization-robust model architectures, which can partially compensate for the lack of hardware acceleration when combined with resolution scaling. Furthermore, the analysis hig ights the importance of hybrid-resolution processing in multi-stage pipelines, where different stages can operate at different input resolutions to balance accuracy and performance. The findings demonstrate that optimal system design requires a joint consideration of hardware characteristics, model architecture, and quantization strategy, rather than relying on a single optimization dimension. The presented results provide practical insights for the design of efficient and robust edge-based ALPR systems, with direct implications for real-world industrial deployments.<\/jats:p>","DOI":"10.3390\/s26165034","type":"journal-article","created":{"date-parts":[[2026,8,10]],"date-time":"2026-08-10T09:24:53Z","timestamp":1786353893000},"page":"5034","update-policy":"https:\/\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Quantized vs. Full-Precision YOLO Models on Edge Devices: A Performance Benchmark for Real-Time License Plate Detection in Smart Parking Systems"],"prefix":"10.3390","volume":"26","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-4381-6555","authenticated-orcid":false,"given":"Ervin","family":"Burkus","sequence":"first","affiliation":[{"name":"Biomatics and Applied Artificial Intelligence Institute, John von Neumann Faculty of Informatics, Obuda University, 1034 Budapest, Hungary"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-6048-3104","authenticated-orcid":false,"given":"Bence","family":"Lesty\u00e1n","sequence":"additional","affiliation":[{"name":"Asura Technologies, 1122 Budapest, Hungary"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-9879-5664","authenticated-orcid":false,"given":"Lehel","family":"D\u00e9nes-Fazakas","sequence":"additional","affiliation":[{"name":"Biomatics and Applied Artificial Intelligence Institute, John von Neumann Faculty of Informatics, Obuda University, 1034 Budapest, Hungary"},{"name":"Physiological Controls Research Center, University Research and Innovation Center, Obuda University, 1034 Budapest, Hungary"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8038-2210","authenticated-orcid":false,"given":"Gy\u00f6rgy","family":"Eigner","sequence":"additional","affiliation":[{"name":"Biomatics and Applied Artificial Intelligence Institute, John von Neumann Faculty of Informatics, Obuda University, 1034 Budapest, Hungary"},{"name":"Physiological Controls Research Center, University Research and Innovation Center, Obuda University, 1034 Budapest, Hungary"},{"name":"MedTech Innovation and Education Center, Obuda University, 1034 Budapest, Hungary"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2026,8,8]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"447","DOI":"10.1109\/TWC.2019.2946140","article-title":"Edge AI: On-Demand Accelerating Deep Neural Network Inference via Edge Computing","volume":"19","author":"Li","year":"2020","journal-title":"IEEE Trans. Wirel. Commun."},{"key":"ref_2","unstructured":"Han, S., Mao, H., and Dally, W.J. (2015). Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding. arXiv."},{"key":"ref_3","doi-asserted-by":"crossref","first-page":"2295","DOI":"10.1109\/JPROC.2017.2761740","article-title":"Efficient Processing of Deep Neural Networks: A Tutorial and Survey","volume":"105","author":"Sze","year":"2017","journal-title":"Proc. IEEE"},{"key":"ref_4","unstructured":"Redmon, J., and Farhadi, A. (2018). YOLOv3: An Incremental Improvement. arXiv."},{"key":"ref_5","unstructured":"Jocher, G., Chaurasia, A., Qiu, J., and Stoken, A. (2022). YOLOv5 by Ultralytics. GitHub Repos., Available online: https:\/\/github.com\/ultralytics\/yolov5\/releases#release-v6.2."},{"key":"ref_6","unstructured":"Ultralytics (2025, September 27). YOLOv8 Docs. Available online: https:\/\/docs.ultralytics.com\/."},{"key":"ref_7","doi-asserted-by":"crossref","first-page":"593","DOI":"10.1007\/978-3-030-01258-8_36","article-title":"License Plate Detection and Recognition in Unconstrained Scenarios","volume":"Volume 11216","author":"Ferrari","year":"2018","journal-title":"Computer Vision\u2013ECCV 2018"},{"key":"ref_8","unstructured":"Tang, X., Han, S., Zhang, L.L., Cao, T., and Liu, Y. (2021, January 8\u201311). To Bridge Neural Network Design and Real-World Performance: A Behaviour Study for Neural Networks. Proceedings of the Machine Learning and Systems 3 (MLSys 2021), Xiamen, China."},{"key":"ref_9","doi-asserted-by":"crossref","first-page":"104712","DOI":"10.1016\/j.jpdc.2023.05.001","article-title":"Exploring Edge TPU for Network Intrusion Detection in IoT","volume":"179","author":"Hosseininoorbin","year":"2023","journal-title":"J. Parallel Distrib. Comput."},{"key":"ref_10","doi-asserted-by":"crossref","first-page":"9102","DOI":"10.1109\/ACCESS.2020.2964608","article-title":"Real-Time Apple Detection System Using Embedded Systems With Hardware Accelerators: An Edge AI Application","volume":"8","author":"Mazzia","year":"2020","journal-title":"IEEE Access"},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"428","DOI":"10.1016\/j.sysarc.2019.01.011","article-title":"A Survey on Optimized Implementation of Deep Learning Models on the NVIDIA Jetson Platform","volume":"97","author":"Mittal","year":"2019","journal-title":"J. Syst. Archit."},{"key":"ref_12","doi-asserted-by":"crossref","unstructured":"Febvay, M., and Bounekkar, A. (2022, January 26\u201327). Deep Learning Frameworks Evaluation for Image Classification on Resource Constrained Device. Proceedings of the 11th International Conference on Embedded Systems and Applications (EMSA 2022), Sydney, Australia.","DOI":"10.5121\/csit.2022.120603"},{"key":"ref_13","doi-asserted-by":"crossref","unstructured":"Mohammadi, M., Smith, H., Khan, L., and Zand, R. (2023, January 5\u20137). Facial Expression Recognition at the Edge: CPU vs. GPU vs. VPU vs. TPU. Proceedings of the ACM Great Lakes Symposium on VLSI, Knoxville, TN, USA.","DOI":"10.1145\/3583781.3590245"},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Baller, S.P., Jindal, A., Chadha, M., and Gerndt, M. (2021). DeepEdgeBench: Benchmarking Deep Neural Networks on Edge Devices. arXiv.","DOI":"10.1109\/IC2E52221.2021.00016"},{"key":"ref_15","unstructured":"Banbury, C., Reddi, V.J., Torelli, P., Holleman, J., Jeffries, N., Kiraly, C., Montino, P., Kanter, D., Ahmed, S., and Pau, D. (2021). MLPerf Tiny Benchmark. arXiv."},{"key":"ref_16","doi-asserted-by":"crossref","unstructured":"Amanatidis, P., Iosifidis, G., and Karampatzakis, D. (2021, January 26\u201328). Comparative Evaluation of Machine Learning Inference Machines on Edge-Class Devices. Proceedings of the 25th Pan-Hellenic Conference on Informatics, Volos, Greece.","DOI":"10.1145\/3503823.3503843"},{"key":"ref_17","doi-asserted-by":"crossref","unstructured":"Howard, A., Sandler, M., Chen, B., Wang, W., Chen, L., Tan, M., Chu, G., Vasudevan, V., Zhu, Y., and Pang, R. (2019). Searching for MobileNetV3. Proceedings of the IEEE\/CVF International Conference on Computer Vision, IEEE.","DOI":"10.1109\/ICCV.2019.00140"},{"key":"ref_18","unstructured":"Raspberry Pi Ltd (2023). Raspberry Pi 5 Product Brief, Raspberry Pi Ltd."},{"key":"ref_19","unstructured":"Axis Communications (2026, July 16). Axis ARTPEC-8 SoC White Paper. Available online: https:\/\/www.axis.com\/search?q=Axis+Communications."},{"key":"ref_20","unstructured":"ARM (2023). Cortex-A76 Technical Reference Manual, ARM."},{"key":"ref_21","unstructured":"ARM (2023). Cortex-A53 Technical Reference Manual, ARM."},{"key":"ref_22","doi-asserted-by":"crossref","unstructured":"Redmon, J., Divvala, S., Girshick, R., and Farhadi, A. (2016, January 27\u201330). You Only Look Once: Unified, Real-Time Object Detection. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA.","DOI":"10.1109\/CVPR.2016.91"},{"key":"ref_23","doi-asserted-by":"crossref","unstructured":"Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C., and Berg, A.C. (2016). SSD: Single Shot MultiBox Detector. Proceedings of the European Conference on Computer Vision, Springer.","DOI":"10.1007\/978-3-319-46448-0_2"},{"key":"ref_24","unstructured":"Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., and Adam, H. (2017). MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications. arXiv."},{"key":"ref_25","doi-asserted-by":"crossref","unstructured":"Huang, J., Rathod, V., Sun, C., Zhu, M., Korattikara, A., Fathi, A., Fischer, I., Wojna, Z., Song, Y., and Guadarrama, S. (2017, January 21\u201326). Speed\/Accuracy Trade-Offs for Modern Convolutional Object Detectors. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA.","DOI":"10.1109\/CVPR.2017.351"},{"key":"ref_26","doi-asserted-by":"crossref","unstructured":"Lin, T.Y., Maire, M., Belongie, S., Hays, J., Perona, P., Ramanan, D., Doll\u00e1r, P., and Zitnick, C.L. (2014). Microsoft COCO: Common Objects in Context. Proceedings of the European Conference on Computer Vision, Springer.","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"ref_27","unstructured":"Jocher, G. (2026, July 16). YOLOv5 by Ultralytics. Available online: https:\/\/github.com\/ultralytics\/yolov5."},{"key":"ref_28","unstructured":"Axis Communications (2026, July 16). YOLOv5 on Axis Cameras. Available online: https:\/\/github.com\/axiscommunications\/axis-model-zoo\/blob\/main\/docs\/yolov5.md."},{"key":"ref_29","unstructured":"Axis Communications (2026, July 16). DLPU Model Conversion. Available online: https:\/\/developer.axis.com\/computer-vision\/computer-vision-on-device\/dlpu-model-conversion\/."},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"102907","DOI":"10.1016\/j.cviu.2020.102907","article-title":"UA-DETRAC: A New Benchmark and Protocol for Multi-Object Detection and Tracking","volume":"93","author":"Wen","year":"2020","journal-title":"Comput. Vis. Image Underst."}],"container-title":["Sensors"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/16\/5034\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,11]],"date-time":"2026-08-11T04:36:52Z","timestamp":1786423012000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.mdpi.com\/1424-8220\/26\/16\/5034"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,8]]},"references-count":30,"journal-issue":{"issue":"16","published-online":{"date-parts":[[2026,8]]}},"alternative-id":["s26165034"],"URL":"https:\/\/doi.org\/10.3390\/s26165034","relation":{},"ISSN":["1424-8220"],"issn-type":[{"value":"1424-8220","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,8]]}}}