Fully convolutional networks for semantic segmentation

[Xin Chen]

      I have been around with the FCN models on the branch of caffe for some days,  but I still don't know what is the meaning of the crop layer, especially its diagonal affine map. Is there a formula about the diagonal affine?

      Thank you for your reply!