I have gone through Deepfashion paper.I have query regarding Landmark pooling layer.
1) Image has 4,6 or 8 landmark points depending on cloth type. So how do you decide size of last FC layer in landmark branch? This size will depend on landmark points but we have variable landmark points.
2) I have eight landmark points for an image and feature maps of conv4. I have to do max pooling around landmark l to get local feature map, similar to Fast-RCNN. But which point on feature maps of conv4 corresponds to landmark point l (x,y). Also what is the kernel size for max pooling?
