Performance Evaluation of Vision Transformers vs. Convolutional Neural Networks for Pavement Crack Detection

As the interest of world is shifting towards the correct image recognition methods in industry, the introduction of deep learning containing the convolutional neural network has changed it all. But recently Vision Transformers have also increased and become better though they were earlier used more in the Natural Language Processing. This paper explores the Efficacy of Vision Transformers vs. CNNs on study based on Pavement Crack Detection, very important in civil infrastructure monitoring. We compare the performances of both models in terms of accuracy, training efficiency, and interpretability with a fixed dataset and by applying both CNN and ViT models to the same tasks. The result gives a look into the pros and cons of each model, providing a useful view for future studies and applications in actual situations and inspection tasks.