Efficient Pothole Segmentation Using Quantization and Pruning for ADAS

William D'Abruzzo Martins; Fernando Santos Osório; Diego Renan Bruno

doi:10.22456/2175-2745.150751

Authors

William D'Abruzzo Martins Universidade Estadual Paulista (Unesp) https://orcid.org/0009-0003-6781-8595
Fernando Santos Osório Universidade de São Paulo (USP) https://orcid.org/0000-0002-6620-2794
Diego Renan Bruno Universidade Estadual Paulista (Unesp)

DOI:

https://doi.org/10.22456/2175-2745.150751

Keywords:

segmentation, ADAS, quantization, model pruning

Abstract

Real-time semantic segmentation is critical for Advanced Driver-Assistance Systems (ADAS), but deploying deep learning models on resource-constrained hardware remains a challenge. This work presents an efficient pipeline for pothole segmentation using the lightweight PP-LiteSeg model, evaluating and comparing two key optimization techniques: post-training quantization and network pruning. Results demonstrate that 8-bit integer (INT8) quantization is highly effective, achieving a 3.54x speedup in GPU compute time with only a minimal, well-controlled loss in segmentation accuracy. While network pruning also reduces model size and latency, it comes at a more pronounced cost to performance. In conclusion, while quantization offers a superior trade-off between speed and accuracy, network pruning remains a valuable strategy for applications where memory compression is the primary objective.

Downloads

Download data is not yet available.

References

[1] SAFYARI, Y.; MAHDIANPARI, M.; SHIRI, H. A review of vision-based pothole detection methods using computer vision and machine learning. Sensors, MDPI AG, v. 24, n. 17, p. 5652, ago. 2024. ISSN 1424-8220. Disponível em: http://dx.doi.org/10.3390/s24175652.

[2] DHIMAN, A.; KLETTE, R. Pothole detection using computer vision and learning. IEEE Transactions on Intelligent Transportation Systems, v. 21, n. 8, p. 3536–3550, 2020.

[3] Confederação Nacional do Transporte (CNT). Radar CNT do Transporte – Pontos Críticos 2023. Brazil, 2023. [CNT Transport Radar – Critical Points 2023]. Available at: https://cnt.org.br/analises-transporte.

[4] MCKENNA, F. P. The human factor in driving accidents an overview of approaches and problems. Ergonomics, Informa UK Limited, v. 25, n. 10, p. 867–877, out. 1982. ISSN 1366-5847. Disponível em: http://dx.doi.org/10.1080/00140138208925047.

[5] LYDON, D. et al. A review of vision-based methods for pothole detection and road profile analysis. In: Civil Engineering Research in Ireland 2020. Cork, Ireland: Cork Institute of Technology, 2020.

[6] MA, N. et al. Computer vision for road imaging and pothole detection: a state-of-the-art review of systems and algorithms. Transportation Safety and Environment, v. 4, n. 4, p. tdac026, 11 2022. ISSN 2631-4428. Disponível em: https://doi.org/10.1093/tse/tdac026.

[7] CHENG, L.; ZHANG, X.; SHEN, J. Road surface condition classification using deep learning. Journal of Visual Communication and Image Representation, Elsevier BV, v. 64, p. 102638, out. 2019. ISSN 1047-3203. Disponível em: http://dx.doi.org/10.1016/j.jvcir.2019.102638.

[8] MA, N. et al. Computer vision for road imaging and pothole detection: a state-of-the-art review of systems and algorithms. Transportation Safety and Environment, Oxford University Press (OUP), v. 4, n. 4, nov. 2022. ISSN 2631-4428. Disponível em: http://dx.doi.org/10.1093/tse/tdac026.

[9] YANG, J. et al. Semantic segmentation for autonomous driving. In: FAN, R.; GUO, S.; BOCUS, M. J. (Ed.). Autonomous Driving Perception: Fundamentals and Applications. Singapore: Springer Nature Singapore, 2023. p. 101–137. ISBN 978-981-99-4287-9. Disponível em: https://doi.org/10.1007/978-981-99-4287-9_4.

[10] MARTINS, W. D.; OSORIO, F. S.; BRUNO, D. R. Real time deep learning based pothole detection on low cost embedded devices for adas. In: 2025 Brazilian Conference on Robotics (CROS). Belo Horizonte, Brazil: IEEE, 2025. v. 1, p. 1–6.

[11] PENG, J. et al. PP-LiteSeg: A Superior Real-Time Semantic Segmentation Model. 2022. Disponível em: https://arxiv.org/abs/2204.02681.

[12] FAN, R. et al. We learn better road pothole detection: From attention aggregation to adversarial domain adaptation. In: Computer Vision – ECCV 2020 Workshops. Springer International Publishing, 2020. p. 285–300. ISBN 9783030668235. Disponível em: http://dx.doi.org/10.1007/978-3-030-66823-5_17.

[13] FAN, R. et al. Graph attention layer evolves semantic segmentation for road pothole detection: A benchmark and algorithms. IEEE Transactions on Image Processing, v. 30, p. 8144–8154, 2021.

[14] TRIPATHI, R.; INDU, S.; KUMAR, R. Ercu-net: segmentation of road potholes using enhanced residual convolutional block based on u-net for adas. Signal, Image and Video Processing, Springer Science and Business Media LLC, v. 18, n. S1, p. 385–394, abr. 2024. ISSN 1863-1711. Disponível em: http://dx.doi.org/10.1007/s11760-024-03161-w.

[15] FENG, Z. et al. Mafnet: Segmentation of road potholes with multimodal attention fusion network for autonomous vehicles. IEEE Transactions on Instrumentation and Measurement, v. 71, p. 1–12, 2022.

[16] AKAGIC, A.; BUZA, E.; OMANOVIC, S. Pothole detection: An efficient vision based method using rgb color space image segmentation. In: 2017 40th International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO). Opatija, Croatia: IEEE, 2017. p. 1104–1109.

[17] MASIHULLAH, S. et al. Attention based coupled framework for road and pothole segmentation. In: 2020 25th International Conference on Pattern Recognition (ICPR). Milan, Italy: IEEE, 2021. p. 5812–5819.

[18] ASAD, M. H. et al. Pothole detection using deep learning: A real-time and ai-on-the-edge perspective. Advances in Civil Engineering, Wiley, v. 2022, n. 1, jan. 2022. ISSN 1687-8094. Disponível em: http://dx.doi.org/10.1155/2022/9221211.

[19] MUSA, A. et al. A lightweight cnn-based pothole detection model for embedded systems using knowledge distillation. IOS Press, 2022. ISSN 1879-8314. Disponível em: http://dx.doi.org/10.3233/FAIA220281.

[20] NIENABER, S.; BOOYSEN, M.; KROON, R. Detecting potholes using simple image processing techniques and real-world footage. In: Proc. 34th Annu. Southern African Transport Conf. Pretoria, South Africa: CSIR International Convention Centre, 2015. p. 1–10.

[21] JAKUBEC, M. et al. Comparison of cnn-based models for pothole detection in real-world adverse conditions: Overview and evaluation. Applied Sciences, v. 13, n. 9, 2023. ISSN 2076-3417. Disponível em: https://www.mdpi.com/2076-3417/13/9/5810.

[22] RAVI, N. et al. Sam 2: Segment anything in images and videos. arXiv preprint, 2024.

[23] ZHAO, J.; ZHANG, Z.; QI, F. An intelligent image segmentation annotation method based on segment anything model. In: CRUZ, C.; ZHANG, Y.; GAO, W. (Ed.). Intelligent Computers, Algorithms, and Applications. Singapore: Springer Nature Singapore, 2024. p. 237–243. ISBN 978-981-97-0065-3.

[24] REZA, M. A. et al. SegBuilder: A Semi-Automatic Annotation Tool for Segmentation. In: 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). Los Alamitos, CA, USA: IEEE Computer Society, 2025. p. 8494–8503. Disponível em: https://doi.ieeecomputersociety.org/10.1109/WACV61041.2025.00823.

[25] MUMUNI, F.; MUMUNI, A. Segment Anything Model for automated image data annotation: empirical studies using text prompts from Grounding DINO. 2024. Disponível em: https://arxiv.org/abs/2406.19057.

[26] ZHAO, H. et al. Pyramid Scene Parsing Network. 2017. Disponível em: https://arxiv.org/abs/1612.01105.

[27] LONG, J.; SHELHAMER, E.; DARRELL, T. Fully convolutional networks for semantic segmentation. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA, USA: IEEE Computer Society, 2015. p. 3431–3440. ISSN 1063-6919. Disponível em: https://doi.ieeecomputersociety.org/10.1109/CVPR.2015.7298965.

[28] SHAFIQUE, M. A.; MUNIR, A.; KONG, J. Deep learning performance characterization on gpus for various quantization frameworks. AI, MDPI AG, v. 4, n. 4, p. 926–948, out. 2023. ISSN 2673-2688. Disponível em: http://dx.doi.org/10.3390/ai4040047.

[29] WU, H. et al. Integer Quantization for Deep Learning Inference: Principles and Empirical Evaluation. 2020. Disponível em: https://arxiv.org/abs/2004.09602.

[30] MCKINSTRY, J. L. et al. Discovering low-precision networks close to full-precision networks for efficient inference. In: 2019 Fifth Workshop on Energy Efficient Machine Learning and Cognitive Computing - NeurIPS Edition (EMC2-NIPS). Vancouver, BC, Canada: IEEE, 2019. p. 6–9.

[31] CUN, Y. L.; DENKER, J. S.; SOLLA, S. A. Optimal brain damage. In: Proceedings of the 3rd International Conference on Neural Information Processing Systems. Cambridge, MA, USA: MIT Press, 1989. (NIPS’89), p. 598–605.

[32] HAN, S. et al. Learning both weights and connections for efficient neural networks. In: Proceedings of the 29th International Conference on Neural Information Processing Systems - Volume 1. Cambridge, MA, USA: MIT Press, 2015. (NIPS’15), p. 1135–1143.

[33] ANWAR, S.; HWANG, K.; SUNG, W. Structured pruning of deep convolutional neural networks. J. Emerg. Technol. Comput. Syst., Association for Computing Machinery, New York, NY, USA, v. 13, n. 3, fev. 2017. ISSN 1550-4832. Disponível em: https://doi.org/10.1145/3005348.

[34] KUZMIN, A. et al. Pruning vs quantization: Which is better? 2023.

[35] FANG, G. et al. DepGraph: Towards Any Structural Pruning. In: 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Los Alamitos, CA, USA: IEEE Computer Society, 2023. p. 16091–16101. Disponível em: https://doi.ieeecomputersociety.org/10.1109/CVPR52729.2023.01544.

[36] FAN, M. et al. Rethinking bisenet for real-time semantic segmentation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Nashville, TN, USA: IEEE, 2021. p. 9716–9725.

[37] AHARON, S. et al. Super-Gradients. GitHub, 2021. Disponível em: https://zenodo.org/record/7789328.

[38] ZHONG, Z. et al. Random Erasing Data Augmentation. 2017. Disponível em: https://arxiv.org/abs/1708.04896.

[39] SANDRU, A.; GEORGESCU, M.-I.; IONESCU, R. T. Feature-level augmentation to improve robustness of deep neural networks to affine transformations. 2022. Disponível em: https://arxiv.org/abs/2202.05152.

[40] LIU, Z. et al. Rethinking the value of network pruning. In: International Conference on Learning Representations. OpenReview, 2019. Disponível em: https://openreview.net/forum?id=rJlnB3C5Ym.