Text this: A multi-class hybrid variational autoencoder and vision transformer model for enhanced plant disease identification