IntelliSign: A Real-Time YOLOv5 based Traffic Sign Detection and Alert System

Abstract

Traffic signs play a vital role in ensuring safe and efficient road usage by providing drivers with regulatory, warning, and informational guidance. However, in real driving conditions, signs are often missed due to distractions, complex traffic scenarios, unfamiliar routes, or challenging environmental factors such as glare, shadows, and poor lighting. Missing critical traffic signs can increase the likelihood of accidents and reduce situational awareness. With the rapid advancement of intelligent transportation systems, the need for automated, real-time traffic sign detection has become significant. IntelliSign, a real-time traffic sign detection and driver alert system developed using the YOLOv5 deep learning architecture. The system is capable of processing images, prerecorded videos, and to accurately detect and classify traffic signs with high confidence. OpenCV is used for real-time frame extraction and preprocessing, while pyttsx3 provides offline text-to-speech alerts to immediately notify users when a sign is detected. A Streamlit-based interface enables intuitive visualization and user interaction, supporting seamless real-time operation. Experimental results demonstrate that IntelliSign performs reliably across varied lighting conditions, achieving low-latency inference suitable for practical deployment. The system offers a scalable, software-only approach to improve driver awareness and enhance road safety applications. systems.

Introduction

The text presents IntelliSign, a real-time traffic sign detection and driver alert system developed to improve road safety and driver awareness. Traffic signs are critical for safe transportation, yet they are often missed due to distractions, unfamiliar routes, high speeds, or poor environmental conditions, contributing to accidents worldwide. With advances in deep learning and computer vision, automated traffic sign detection has become feasible, but many existing systems depend on expensive hardware, cloud connectivity, or limited deployment environments, restricting their practical use.

IntelliSign addresses these limitations by providing a lightweight, accessible, and offline-capable solution. It integrates YOLOv5 for fast and accurate traffic sign detection, OpenCV for real-time video processing, Streamlit for user interaction and visualization, and pyttsx3 for offline voice alerts. The system is designed to run on standard laptops, making it suitable for diverse environments, including low-connectivity or rural areas.

The methodology follows an end-to-end pipeline that includes dataset preparation using an Indian traffic sign dataset, training a YOLOv5s model, real-time frame processing, detection with non-maximum suppression, and audio alert generation. The system ensures that alerts are only triggered for newly detected signs to avoid redundancy. A modular architecture supports efficient input handling, preprocessing, detection, post-processing, visualization, and audio notification.

Experimental results demonstrate strong detection performance on both static images and video streams, stable real-time operation, and responsive audio alerts. While extreme lighting and distant signs remain challenging, overall performance confirms IntelliSign’s suitability for real-world traffic sign awareness. The system offers a practical, user-friendly, and deployable solution that enhances driver safety and supports intelligent transportation applications.

Conclusion

The IntelliSign system was developed with the intention of providing a practical, software-based solution for detecting Indian traffic signs in real time. Through the combination of a YOLOv5 detection engine, a lightweight audio alert module, and an accessible Streamlit interface, the system demonstrates how modern computer vision techniques can be applied to everyday driving-awareness scenarios without relying on specialized hardware. The evaluation results indicate that IntelliSign handles common variations in scene conditions and input modes reasonably well. Signs are detected promptly, visual feedback is clear, and the audio notifications support timely understanding of what appears in the environment. While the system performs effectively in standard usage situations, some limitations remain. Scenes with harsh lighting, extreme glare, or very distant signs occasionally reduced confidence or delayed detection. Addressing these cases may require further fine-tuning, such as training with additional samples or applying preprocessing techniques suited to low-contrast environments. Despite these challenges, IntelliSign serves as a strong demonstration of how a compact and easily deployable detection pipeline can contribute to improving driver awareness. It provides a foundation that can be extended with future enhancements such as multi-language alerts, broader sign coverage, or integration with navigation tools.

References

[1] M. B. Jensen, K. Nasrollahi and T. B. Moeslund, “Evaluating state-of-the-art object detectors on challenging traffic light data,” Proc. IEEE CVPR Workshops, pp. 9–15, 2017. [2] A. Santos, P. A. R. Abu, C. Oppus and R. S. J. Reyes, “Real-time traffic sign detection and recognition system for assistive driving,” Ateneo Research Archive, 2020. [3] S. Huang, H. Lin and C. Chang, “An in-car camera system for traffic sign detection and recognition,” in Proc. IFSA–SCIS, pp. 1–6, 2017. [4] Z. Bi, L. Yu, H. Gao et al., “Improved VGG model-based efficient traffic sign recognition for safe driving in 5G scenarios,” Int. J. Mach. Learn. Cybern., 2020. [5] M. Karaduman and H. Eren, “Deep learning-based traffic direction sign detection and determining driving style,” in Proc. Int. Conf. Computer Science and Engineering (UBMK), pp. 1046–1050, 2017. [6] World Health Organization, “Road traffic injuries,” Fact Sheet, Mar. 2022. [7] Ultralytics, “YOLOv5: PyTorch implementation,” [8] Kaggle, “Traffic Sign Localization and Detection (SSD-Annotated) Dataset.” [9] J. Redmon, S. Divvala, R. Girshick and A. Farhadi, “You Only Look Once: Unified, real-time object detection,” in Proc. IEEE CVPR, pp. 779–788, 2016. [10] A. Bochkovskiy, C.-Y. Wang and H.-Y. M. Liao, “YOLOv4: Optimal speed and accuracy of object detection,” arXiv preprint, arXiv:2004.10934, 2020. [11] K. He, X. Zhang, S. Ren and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE CVPR, 2016 [12] W. Liu et al., “SSD: Single shot multibox detector,” in Proc. ECCV, 2016. [13] M. Everingham et al., “The Pascal Visual Object Classes (VOC) Challenge,” Int. J. Comput. Vision, vol. 88, pp. 303–338, 2010. [14] T.-Y. Lin et al., “Feature Pyramid Networks for object detection,” in Proc. IEEE CVPR, 2017 [15] M. Tan and Q. Le, “EfficientNet: Rethinking model scaling for convolutional neural networks,” in Proc. ICML, 2019. [16] X. Wang et al., “A survey of YOLO-based object detection systems and future trends,” arXiv preprint, arXiv:2104.06066, 2021. [17] “Advances in vision-based intelligent transportation systems,” Pattern Recognition Letters, Elsevier, 2020. [18] “Computer vision techniques for road safety analysis,” Machine Vision and Applications, Springer, 2019. [19] “Real-time visual perception models for intelligent vehicles,” in Proc. IEEE Intelligent Transportation Systems Conference, IEEE ITS Society, 2018.

Copyright

Copyright © 2025 Harshitha DR, Shalini J, Sneha V, Varshitha N, Mrs. Divya S, Dr. Vidyarani H J. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.