Should fooling_rate exclude true_label != est_labels_pert ?

[543877815]

May I ask, when caculate

fooling_rate = float(np.sum(est_labels_pert != est_labels_orig) / float(num_images))

Should fooling_rate exclude sth like true_label != est_labels_pert ?
Because when true_label != est_labels_pert, we don't need to fool the network, I think.