Enhancing Classification of Traffic Sign using Multi-Technique Data Augmentation

Intelligent Transportation Systems (ITS) plays a crucial role in improving road safety and efficiency while improving environment and lowering pollution. One important aspect of ITS is the accurate classification of traffic signs, which aids in automated decision-making processes. Deep learning (DL) have shown remarkable performance in traffic sign classification. However, due to the limited size of available training datasets, these models often face challenges in generalizing to real-world scenarios. In this research paper, we propose a novel approach to enhance the accuracy of traffic sign classification in ITS by leveraging more than 20 advanced multi-technique data augmentation methods that have not been explored in pre-existing literature. This multi-technique significantly improves the performance of deep learning models by applying controlled transformations to the existing dataset, increasing the diversity and robustness of the training data. By leveraging convolutional neural networks (CNNs) and fine-tuning them with the augmented training dataset, we empower the models to learn meaningful representations. Through a comparative analysis with baseline models trained on the original dataset, which used common data augmentations in previous studies, we demonstrate the performance improvement achieved by our proposed approach. The results demonstrate importance and effectiveness of using multi-technique data augmentation in enhancing the accuracy of traffic sign classification within the context of ITS.