Advanced Traffic Sign Detection Using a Hybrid Deep Learning Architecture with Enhanced Dense Layer Customization

Accurate traffic sign detection is crucial for autonomous driving systems to ensure road safety and enable reliable decision-making. Traditional models like ResNet101 and Inception V3 offer strong individual performance, yet each has limitations in depth and multi-scale feature handling. In this paper, we propose a fusion model that combines ResNet101 and Inception V3, leveraging deep feature extraction and multi-scale feature processing. Using the ICTS dataset, our model achieves an accuracy of 98.58 %, significantly outperforming single-architecture models. This fusion approach presents a robust and efficient solution for real-time traffic sign detection in autonomous vehicle systems.