Upsampling output?
sabrinazuraimi opened this issue · comments
Hi, I'm trying to use the output of the network to move a few certain points(e.g: at pixel 441) however the output of the network is 1/16 of the input size.
My question is it possible to upsample/deconvolve the output so that I get a flow map that is the same size as the input image? (Or is there any other way to see how an exact pixel moved in the downsampled flow map?)
Yes of course you can!
You might want to see the torch.nn.functional.interpolate function that will get you sorted out.
https://pytorch.org/docs/stable/nn.functional.html#interpolate