FECT

Abstract: This researcher only focuses on five classes, namely mikrolet lanes one, six, eight, ten, and eleven. Here, mikrolet vehicle path features are automatically extracted and categorized. During the analysis, various metrics were used for evaluation, including accuracy, precision, recall, and F1-Score. The proposed model achieves progressive accuracy based on vehicle characteristics. There are three pre-trained deep learning models of seizure neural networks, AlexNet, VGG-16, and ResNet-50. Based on the comparison of the results, in the mikrolet vehicle path classification, our recommended ResNet-50 pre-training model achieves an accuracy of 100%.

Keywords: AlexNet, CNN, Deep-Learning, Mikrolet, Multi-class classification, ResNet-50, Vehicle, VGG-16.