Deformation Layer
In this subsection we would like to explain what happens
in the second convolutional layer. This layer receives the P
part detection maps as input and part scores s = fs1; :::; sP g,
P = 8, are its output. The convolutional layer treats the detection
maps individually and produces the pth part score
sp from the pth part detection map, denoted by Mp. A 2D
summation map, Bp, is calculated by summing up the part
detection map Mp and the deformation maps as follows: