Skip to main content

Table 1 U-Net architecture used for the DCGMM in our study

From: An automatic nuclei segmentation method based on deep convolutional neural networks for histopathology images

LayerDetails LayerDetails
Input  Output 
   
  conv9_31x1x32; ReLU
  conv9_23x3x64; ReLU
conv1_13x3x32; ReLU conv9_13x3x64; ReLU
conv1_23x3x32; ReLUconcat4concatenate upsample4 with conv1_2
   
pool12x2 max pool stride 2  
  upsample42x2 upsample of conv8
conv2_13x3x64; ReLU conv83x3x32; ReLU
conv2_23x3x64;ReLUconcat3concatenate upsample3 with conv2_2
   
pool22x2 max pool stride 2  
  upsample32x2 upsample of conv7
conv3_13x3x128; ReLU conv73x3x64; ReLU
conv3_21x1x128; ReLU  
conv3_31x1x128; ReLUconcat2concatenate upsample2 with conv3_3
   
pool32x2 max pool stride 2  
conv4_13x3x256; ReLU upsample22x2 upsample of conv6
conv4_23x3x256; ReLU conv63x3x128; ReLU
conv4_31x1x256;ReLUconcat1concatenate upsample1 with conv4_3
   
pool42x2 max pool stride 2  
conv5_13x3x256; ReLU upsample12x2 upsample of conv5_3
conv5_23x3x256; ReLU  
conv5_31x1x256; ReLU