Issuu

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395-0056

Volume: 11 Issue: 05 | May 2024

p-ISSN: 2395-0072

www.irjet.net

Traffic Sign Detection and Classification Using Machine Learning Ashish Mahur1, Akash Gupta2, Ms. Geeta3 Student, Dept of CSE-AI, NIET Greater Noida, Uttar Pradesh, India, Student, Dept of CSE-AI, NIET Greater Noida, Uttar Pradesh, India, Assistant Professor, Dept Of CSE-AI, NIET Greater Noida, Uttar Pradesh, India. ---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Traffic sign classification and recognition are main components of ADAS and self-driving vehicles. In this research, we suggested a fresh approach for efficient signature recognition using the Mobile Net v1 convolutional neural network (CNN). The German vehicle signature for recognition of test data is used for training and evaluation purposes. Our approach focuses on using the light source of Mobile Net v1 to achieve high accuracy in traffic tag classification while maintaining performance, making it suitable for applications of the time. We first processed the data to improve its applicability to the Mobile Net v1 architecture to ensure efficient use of computing resources without sacrificing performance. By means of comprehensive experiments, We exhibit the efficacy of our methodology, compared to traditional CNN architectures. Our model achieves competitive accuracy while minimizing operational complexity and memory footprint, making it ideal for deployment in constrained environments such as embedded devices and mobile devices. Key Words: Traffic sign classification, TSD, TSR, CNN, TensorFlow, colab, Traffic sign, GTSRB Dataset.

1.INTRODUCTION Although traffic technology is developing rapidly, traffic safety is still very important. Traffic Sign Classification (TSC) systems play an important role in improving traffic safety through automatic recognition and interpretation of traffic signs photos, navigation devices, hazard avoidance and vehicle automation. The source of the study is the GTSR dataset, which is a database containing traffic sign photos. [4] In this research paper, we embark on a journey to unravel the intricacies of the GTSRB dataset through the lens of preprocessing and visualization. Our endeavour is to illuminate the path for researchers and practitioners seeking to harness the potential of this dataset in advancing TSC technology.[11][12] Our methodology encompasses a meticulous process of data acquisition, preprocessing, and visualization. We delve into the challenges posed by diverse image formats, the intricacies of labelling traffic sign categories, and the insights gleaned from visual exploration. Beyond mere technicalities, this paper delves into the broader implications of TSC research. We explore how advancements in machine learning, computer vision, and sensor technologies converge to redefine the landscape of road safety and transportation systems. Moreover, we shed light on the societal impact of TSC technology, from improving accessibility for drivers with disabilities to enhancing the efficiency of urban traffic management.[14][15] As we navigate through the methodologies and discoveries presented in this paper, we invite readers to envision a future where roads are safer, navigation is seamless, and transportation systems are more sustainable. Let us embark on this journey together, where innovation meets responsibility, and the possibilities of TSC technology.

2. LITERATURE REVIEW

Table 1. Literature Review Table

S.NO

Author Year

Technology

Limitation

Zhang, Y., Zhang, L., & Du, Y. (2016)

Convolutional Neural Networks

Requires significant Effective traffic sign recognition computational resources

Sermanet, P., LeCun, Y., Multi-Scale Convolutional High computational & Perronnin, F. (2011) Networks complexity

Multi-scale approach improves accuracy

Houben, S., Stallkamp, J., German Traffic Sign Salmen, J., & Schlipsing, Detection Benchmark M. (2013)

Real-world benchmark for detection

Impact Factor value: 8.226

Dataset-specific performance

Remark

ISO 9001:2008 Certified Journal

Page 1482