An Empirical Analysis of Vision Transformer and CNN in Resource-Constrained Federated Learning.