Questions about DVN network structure for Weizmann horses

[nkg114mc]

I am trying to replicate the experiments of DVN for image segmentation on Weizmann horse. I have some confusion about the network structure on Figure 2 in the paper.

1. How is the last convolution layer and the first fully connected layer connected?
   The paper said that there are three convolution layers, but on the output of “6x6”, there is still a “5x5 convolution kernel” with stride of 2. Is it a fourth convolution layer? Or a pooling? What exactly the input dimension of the first fully connected layer? Is it 6x6x128? Or 3x3x128?

2. For the input, in Figure 2, there are 3+k channels. The 3 channels are for the input image, k is for the musk. What does k represents here? Is it related to the number of classes? For example, for Weizmann horse, is k = 1 (0 as background and 1 for horse)?

Thanks!