Publication

Abstract : For autonomous driving technologies to work effectively and increase road safety, traffic sign detection is essential. The use of transfer learning and customised convolutional neural networks (CNNs) for this crucial task is examined in this paper. As a starting point for transfer learning, we use the pre-trained VGG16 model and evaluate its performance against a CNN architecture that we created. Data preprocessing and data gathering are two aspects of our research that entail a large dataset of traffic signs. We employ strategies for data augmentation to improve the model's generalisation capabilities. Both models exhibit tremendous potential in our studies. While the bespoke CNN architecture succeeds in achieving amazing accuracy, the transfer learning strategy based on VGG16 shows its capacity to produce competitive outcomes. These models' advantages and disadvantages are clarified by our thorough examination and comparison. This study has implications for improving traffic sign detecting systems and laying a strong foundation for practical applications. Our findings show the potential of using pretrained models and creating unique CNNs, paving the path for safer and more effective navigation systems for vehicles. These models' advantages and disadvantages are clarified by our thorough examination and comparison. This study has implications for improving traffic sign detecting systems and laying a strong foundation for practical applications. Our findings show the potential of using pre-trained models and creating unique CNNs, paving the path for safer and more effective navigation systems for vehicles. These models' advantages and disadvantages are clarified by our thorough examination and comparison. This study has implications for improving traffic sign detecting systems and laying a strong foundation for practical applications. Our findings show the potential of using pre-trained models and creating unique CNNs, paving the path for safer and more effective navigation systems for vehicles.