When I was reading this this paper, Fully Convolutional Networks for Semantic Segmentation, I found that they use an up-sampling layer to classify each pixel in to a class. I have two questions:
How do you understand the mathematics behind the de-convolution operation?
Why do we use an upsampling layer? Is it for extract more global context?