Skip to main content
. 2022 Aug 6;34(24):21927–21938. doi: 10.1007/s00521-022-07653-z

Table 2.

Layers of the pre-trained models utilized in the U-Net encoder part

VGG16 ResNet101 DenseNet121 (layer id) InceptionV3 (layer id) EfficientNetB5
block5_conv3 stage4_unit1_relu1 311 228 block6a_expand_activation
block4_conv3 stage3_unit1_relu1 139 86 block4a_expand_activation
block3_conv3 stage2_unit1_relu1 51 16 block3a_expand_activation
block2_conv2 relu0 4 9 block2a_expand_activation