Hi Kobe,
I understand that you want to build a nested UNet or UNet3+ along with a full-scale skip connection.
For building a UNet, MATLAB provides the “unetLayers” function directly, which is a part of the Computer Vision Toolbox. But for nested UNet or UNet3+, you need to build the model from scratch.
For a nested UNet, you can create a UNet normally using the “unetLayers” function and perform modifications on it by extracting the different encoder and decoder layers from it, nesting these layers, and merging the nested encoded and decoded layers. Skip connections can also be added in a similar way.
For example, to extract the layers, you can do the following:
unetModel = unetLayers(inputSize, numClasses);
layers = unetModel.Layers;
encoderLayers = layers(2:18);
decoderLayers = layers(19:end);
bridgeLayer = decoderLayers(1);
decoderLayers = decoderLayers(2:end);
Then, other steps like adding nesting or skip connections can be performed using these extracted layers.
Refer the following link to learn more about "unetLayers" function:
Refer the following example for using “encoderDecoderNetwork” to create a UNet network with skip connections:
Hope this helps!
