Advanced Traffic Dataset for Real-Time Classification Using Data Augmentation

Abstract

For autonomous driving systems to function effectively and ensure safety, traffic sign recognition is significant. The goal of this study is to improve variety and coverage by creating a hybrid dataset for Egyptian traffic signs that combines synthetic and real-world representations. LeNet-5, VGG16, and Inception-V3 are three instances of deep learning models based on CNNs that were used to validate the dataset's performance. Each model was assessed for accuracy and real-time performance. According to the results, computerized traffic sign identification and recognition is crucial for minimizing errors of humanity and adjusting to difficult situations like changing illumination and occlusion. This research contributes to safer and more intelligent transportation by presenting a strong framework for improving autonomous driving in the Egyptian setting.