Regarding the visualization of different network feature maps in the article

[PMRS-lab]

May I ask which part of the network outputs the visualization results of the feature map using different network architectures with random initial weights in the article? Is it an output feature map or a PCA dimensionality reduction visualization display of the features?