Argument

Improved algorithm in this paper, ResNet-50, GoogleNet, MobileNetv1, VGG16

Learning rate

0.001

optimizer

Adam

Loss function

Categorical

Batch size

8

Training times

15