Classification of ancient coins in archaeology using a novel deep learning approach: Bayesian convolutional neural network

This study looks at classifying and dating old coins. Coins are important in archaeology because they tell us about history, culture, and economy. Knowing the right date of coins helps to understand excavation sites and also helps studies in art, politics, and social life. Normally, numismatics experts do this work, but it takes a lot of time and their judgment can be different. In this research, we used some deep learning models like DenseNet-201, GoogLeNet, InceptionV3, MobileNetV2, and Xception. We also tested a new model called Bayesian Convolutional Neural Network (B-CNN). This model uses Bayesian optimization to choose parameters. The B-CNN reached about 97% accuracy, which is better than the other models. The results show that B-CNN can be a good tool for archaeologists, especially for dating coins. It gives more clear and correct results and reduces the need for special experts. The new part of this study is mixing Bayesian optimization with CNNs. This makes the model stronger than older methods. The work connects archaeology and computer science and shows better sensitivity and performance, but it also needs more training time.