Enhancing Road Safety through Hybrid CNN Models: Ensemble Framework for German Traffic Sign Recognition Benchmark (GTSRB)

DOI

10.2991/978-94-6239-664-7_49How to use a DOI?

Keywords

Traffic Sign Recognition; ADAS; GTSRB; LeNet-5; ResNet18; MobileNetV2; Deep Learning; Convolutional Neural Networks (CNN); and Ensemble Learning

Abstract

Traffic Sign Recognition (TSR) is an important part of the Advanced Driver Assistance Systems (ADAS) to guarantee intelligent vehicle safety. Correct interpretation of traffic signs can thus make humans response less prone to mistake, preventing accidents and in general increase the safety of traffic. But real-world issues such as varying illuminations, occlusions and sign distortions frequently preclude traditional machine learning methods. We drew upon the challenging German Traffic Sign Recognition Benchmark (GTSRB) dataset to develop stable and efficient deep learning-based TSR models. We designed and compared four CNNs: LeNet-5, ResNet18, MobileNetV2 and a hybrid ensemble model that combined the predictions of LeNet-5, ResNet18 and MobileNetV2. The fusion layer takes the feature representations of Wildest Fusion from LeNet-5 and MobileNetV2 for an inference at some layers to combine their output using a weighted summation, and mixes it with that of ResNet18 by means of a weighted averaging. This procedure helps to make the most use of the complementary capacity of the two networks. The hybrid ensemble model outperformed the single LeNet-5 (98.21%), ResNet18 (91.71%), and MobileNetV2 (96.65%) to obtain 99% accuracy, respectively. The findings point to the necessity of integrating lightweight and deep architecture for better recognition performance, particularly under challenging low-light conditions (glare and light direction). This work has important implications for intelligent transportation systems and autonomous driving.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Md Muhasin Ali
AU  - Shovan Samanta Turzo
AU  - Antony Tony Mondal
AU  - Nasim Parvez
AU  - Hossain Mohammad Shuvo
AU  - Rifat Bin Saleh
AU  - Md Jahidul Islam Mozumder
AU  - Mohammad Soad Khan
PY  - 2026
DA  - 2026/06/08
TI  - Enhancing Road Safety through Hybrid CNN Models: Ensemble Framework for German Traffic Sign Recognition Benchmark (GTSRB)
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 701
EP  - 715
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_49
DO  - 10.2991/978-94-6239-664-7_49
ID  - Ali2026
ER  -