unet upsample-concatenate problem

[Zakiyi]

Hello, thx for your sharing, I am new to segmentation with deep learning. I have checked the unet paper, in the decode part of unet, the upsampled feature maps seems reduced channel and then concatenated with corresponding feature from encoder part. hence, the channel doubled. however, in your code, the channel of upsampled feature (bilinear) directly concatenated with the encoder part feature, and then passed through unetConv2(two layers convolutional layers), so i wonder isn't wrong here? Supposing upsampled feature channel is 1024, the counterpart feature channel is 512, then the channel of concatenated feature should be 1536, thus it's conflicted with filters[4]=1024 "self.up_concat4 = unetUp(filters[4], filters[3])" filters[4]=1024 I am puzzled :(