I note in the repo, all the torch.nn.functional.interpolate function use the args antialias=True.
I wondered how much the setting will influence the ultimate results compared to use the normal upsample handle torch.nn.functional.interpolate(image_orig_tensor, size=(size, size), mode='bilinear', align_corners=False).
Thanks for providing such impressive work!
I note in the repo, all the
torch.nn.functional.interpolatefunction use the argsantialias=True. I wondered how much the setting will influence the ultimate results compared to use the normal upsample handletorch.nn.functional.interpolate(image_orig_tensor, size=(size, size), mode='bilinear', align_corners=False).Thanks in advance!