When I use your code , i can't get your performance,Can you help me?

[BOBKINGS1101]

I didnt change the train code,use the model "checkpoint = ModelCheckpoint(save_path, monitor='final_out_loss', verbose=1, save_best_only=True, mode='min')",iteration=3, DATASET='STARE',batch_size=32, epochs=200.
predict(batch_size=32, epochs=200, iteration=3, stride_size=3, DATASET='STARE'),and
and result is lower than your paper.
Area under the ROC curve: 0.962542154639597
Area under Precision-Recall curve: 0.8794006909319013
Jaccard similarity score: 0.9677991055713587
F1 score (F-measure): 0.7959445466207131

Confusion matrix:[[1113200 12113]
[ 27496 77250]]
ACCURACY: 0.9677991055713587
SENSITIVITY: 0.7374983292918107
SPECIFICITY: 0.9892358837052446
PRECISION: 0.8644517305820082
Can you help me??where i did in wrong way?

[conscienceli]

I don't know... What's your dataset config. How many images for training/test? Well, 0.96 is impossible in my environment. Even common UNet can get an AUC>0.98.