Deep Learning Fusion Ensemble for Enhanced Traffic Sign Detection Using the ICTS Dataset

Traffic Sign Identification and Detection(TSID) sys- tems are vital components of advanced and smart driving systems as well as autonomous vehicles. Our proposed model, based on deep learning, specifically trains and evaluates convolutional neural networks (CNN) using ensemble pre-trained models such as VGG16 and ResNet50 from the field of transfer learning for image classification and combines them in a fusion ensemble model. We leveraged data augmentation strategies to increase the training images during preprocessing. Our data and outcomes show that the fusion of the proposed model performed better, achieving a classification accuracy of 99.70 % using the ICTS (Indian Cautionary Traffic Sign) dataset compared to some traditional models.