Robust Road Sign Feature Extraction Through Data Curation and Multi-Task Learning for Global Map ‎Creation

  • Authors

    • Godson D’silva Principal Data Scientist, Here Technologies, Mumbai, Maharashtra 400708 and PhD Research Scholar, Department of Information Technology, Finolex Academy of Management and Technology, Ratnagiri, ‎Maharashtra 415639
    • Dr. Vinayak Ashok Bharadi Professor & HOD, Department of Information Technology, Finolex Academy of Management and Technology, Ratnagiri, ‎Maharashtra 415639‎
    https://doi.org/10.14419/enqwz712

    Received date: June 27, 2025

    Accepted date: August 5, 2025

    Published date: August 12, 2025

  • Road Sign; Feature Extraction; Data Curation; YOLOv7; Multi-Task Learning; Global Map

  • Abstract

    This research presents a comprehensive approach to enhancing road sign feature extraction for global map creation through ‎strategic improvements in data quality, feature learning, and network architecture. Designed to address core challenges in HERE ‎Technologies' map creation pipeline (US patent PAN: 18/988231), our approach significantly improves the Stage 1 component ‎of their patented three-stage framework by replacing the previous YOLOv7-based implementation with a more robust and ‎effective solution. The methodology centers on three key innovations: (1) an intelligent data curation and filtering strategy that ‎reduces annotation noise by 37% and improves overall data quality without extensive manual re-annotation; (2) novel self-‎supervised pretext tasks that develop rich feature representations of road sign characteristics such as color, shape, and ‎contextual positioning; and (3) a multi-headed network architecture that preserves geometric understanding while enabling ‎simultaneous optimization of detection, segmentation, and classification tasks. These innovations collectively address critical ‎map creation challenges, including domain divergence between different imagery sources, class imbalance across sign types, ‎data scarcity for rare classes, and noisy training samples. Evaluation metrics demonstrate exceptional improvements, with the ‎enhanced system achieving 92% precision, 93% mAP@0.5 for detection, and processing inputs 64.29% faster than the ‎previous implementation while simultaneously performing multiple tasks. The approach significantly improves performance in ‎challenging scenarios, with a 53% improvement in adverse lighting conditions and 31% higher accuracy in poor weather. By ‎focusing on fundamental improvements in data quality, feature representation, and architectural design rather than simply ‎adopting newer base models, this work establishes a foundation for more efficient and accurate feature extraction that enables ‎faster global expansion of map coverage without sacrificing quality‎.

  • References

    1. Behrendt, K., L. Novak, and R. Botros. 2017. “A deep learning approach to traffic lights:Detection, tracking, and classification.” IEEE Int. Conf. Robot. Autom., 1370–1377.Singapore. https://doi.org/10.1109/ICRA.2017.7989163.
    2. Du, L., W. Chen, S. Fu, H. Kong, C. Li, and Z. Pei. 2019. “Real-time detection of vehicle and traffic light for intelligent and connected vehicles based on YOLOv3 network.” 5th Int. Conf.Transp. Inf. Saf., 388–392. Liverpool, UK. https://doi.org/10.1109/ICTIS.2019.8883761.
    3. ensen, M. B., K. Nasrollahi, and T. B. Moeslund. 2017. “Evaluating State-of-the-art Object Detector on Challenging Traffic Light Data.” IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work., 882–888. https://doi.org/10.1109/CVPRW.2017.122.
    4. Jensen, M. B., M. P. Philipsen, A. Møgelmose, T. B. Moeslund, and M. M. Trivedi. 2016.“Vision for Looking at Traffic Lights: Issues, Survey, and Perspectives.” IEEE Trans. Intell.Transp. Syst., 17 (7): 1800–1815. https://doi.org/10.1109/TITS.2015.2509509.
    5. Jocher, G. n.d. “GitHub - ultralytics/yolov5.” Accessed June 1, 2022.https://github.com/ultralytics/yolo.
    6. Li, Z., Q. Zeng, Y. Liu, J. Liu, and L. Li. 2021. “An improved traffic lights recognition algorithm for autonomous driving in complex scenarios.” Int. J. Distrib. Sens. Networks, 17 (5). https://doi.org/10.1177/15501477211018374.
    7. Liu, S., L. Qi, H. Qin, J. Shi, and J. Jia. 2018. “Path Aggregation Network for InstanceSegmentation.” Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit., 8759–8768. https://doi.org/10.1109/CVPR.2018.00913.
    8. Peng, J., M. Xu, and Y. Yan. 2021. “Automatic Recognition of Pointer Meter Reading Based on Yolov4 and Improved U-net Algorithm.” IEEE Int. Conf. Electron. Technol. Commun. Inf.,52–57. Changchun, China. https://doi.org/10.1109/ICETCI53161.2021.9563496.
    9. Possatti, L. C., R. Guidolini, V. B. Cardoso, R. F. Berriel, T. M. Paixão, C. Badue, A. F. De Souza, and T. Oliveira-Santos. 2019. “Traffic Light Recognition Using Deep Learning andPrior Maps for Autonomous Cars.” Int. Jt. Conf. Neural Networks (IJCNN). Budapest, Hungary. https://doi.org/10.1109/IJCNN.2019.8851927.
    10. Wang, C. Y., H. Y. Mark Liao, Y. H. Wu, P. Y. Chen, J. W. Hsieh, and I. H. Yeh. 2019.“CSPNet: A New Backbone that can Enhance Learning Capability of CNN.” IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. Work., 1571–1580. IEEE Computer Society. https://doi.org/10.1109/CVPRW50498.2020.00203.
    11. Wang, Q., Q. Zhang, X. Liang, Y. Wang, C. Zhou, and V. I. Mikulovich. 2022. “Traffic Lights Detection and Recognition Method Based on the Improved YOLOv4 Algorithm.” Sensors,22 (200). https://doi.org/10.3390/s22010200.
    12. Yan, S., X. Liu, W. Qian, and Q. Chen. 2021. “An End-to-End Traffic Light Detection Algorithm Based on Deep Learning.” Int. Conf. Secur. Pat-tern Anal. Cybern., 370–373. https://doi.org/10.1109/SPAC53836.2021.9539934.
    13. Guo, J., You, R., Huang, L., 2020. Mixed vertical-and-horizontal-text traffic sign detection and recognition for street-level scene,2020. IEEE Ac-cess 8, 69413–69425. https://doi.org/10.1109/ACCESS.2020.2986500.
    14. Zhou, S., Qiu, J., 2021. Enhanced SSD with interactive multi-scale attention features for object detection. Multimed. Tools Appl. 80, 11539–11556. https://doi.org/10.1007/s11042-020-10191-2.
    15. Zhang, J., Sun, J., Wang, J., Yue, X.-G., 2021. Visual object tracking based on residual network and cascaded correlation filters. J. Ambient Intell. Human Comput. 12 (8), 8427–8440. https://doi.org/10.1007/s12652-020-02572-0.
    16. Temel, D., Chen, M.H., AlRegib, G., 2020. Traffic sign detection under challenging conditions: a deeper look into performance variations and spec-tral characteristics. IEEE Trans. Intell. Transp. Syst. 21 (9), 3663–3673. https://doi.org/10.1109/TITS.2019.2931429.
    17. Kamal, U., Tonmoy, T.I., Das, S., Hasan, M.K., 2019. Automatic traffic sign detection and recognition using SegU-Net and a modified Tversky loss function with L1-constraint. IEEE Trans. Intell. Transp. Syst. 1–13. https://doi.org/10.1109/TITS.2019.2911727.
    18. Wong, A., Shafiee, M.J., St. Jules, M., 2018. MicronNet: a highly compact deep convolutional neural network architecture for real-time embedded traffic sign classification. IEEE Access 6, 59803–59810. https://doi.org/10.1109/ACCESS.2018.2873948.
    19. Avramovic´, A., Sluga, D., Tabernik, D., Skocˇaj, D., Stojnic´, V., Ilc, N., 2020. Neural-network-based traffic sign detection and recognition in high-definition images using region focusing and parallelization. IEEE Access 8, 189855–189868. https://doi.org/10.1109/ACCESS.2020.3031191.
    20. Lee, H.S., Kim, K., 2018. Simultaneous traffic sign detection and boundary estimation using convolutional neural network. IEEE Trans. Intell. Transp. Syst. 19(5), 1652–1663. https://doi.org/10.1109/TITS.2018.2801560.
    21. G.Dsilva,A.Bhoir,X.Jin and K.Gao, "Systems and Methods for Road Sign Detection from Street Level Imagery Using a Multi-Stage Neural Net-work," U.S. Patent Application 18/988,231, Dec. 19, 2024.

  • Downloads

  • How to Cite

    D’silva , G. ., & Bharadi , D. V. A. . (2025). Robust Road Sign Feature Extraction Through Data Curation and Multi-Task Learning for Global Map ‎Creation. International Journal of Basic and Applied Sciences, 14(4), 368-377. https://doi.org/10.14419/enqwz712