Table 3 Hyper parameters for network training.

image_size

256

num_channels

1

num_layers

2

image_size

256*256

hidden_dim

768

input_dim

2

mlp_dim

3072

output_dim

2

num_heads

8

Learning _rate

0.0007

dropout_rate

0.2

optimizer

Adam

patch_size

16

Loss function

Weighted cross-entropy

num_patches

16*16

Epochs(IWS)

50

Epochs(ICH)

150