Hi, once your model is trained, saved and loaded,you could use normal predict function or test generator function depending upon your testing dataset size and your memory, and for findings the associate class indices,you could save the train_generator.class_indices value in pickle format and load it or just write that to a text file and read it whenever you want.