ChunhuanLin/deform_conv_pytorch

why use mask in the zero padding area?

Closed this issue · 1 comments

In the padding area, the pointer p is calculate as p = p*(1-mask) + floor_p*mask, then on the left side, the bilinear interpolated result is always 0, but on the right side, it's not always 0. This seems a little strange.

Hi, I just find this way will give the same results to which given by MXNet.