What is the effect of Torch.nn.Upsample VS F.resize

1

Hi,

Just curious to know if there is a difference in terms of the pixel value or any if I use torch.nn.upsample vs F.resize? assuming both functions resize/upsample to the same pixel size and the resampling method is Nearest.

Thank you!

2

They are the same inside.
When you look inside the source code of torch.nn.upsample, you will find that the forward function of upsample is relized by F.resize( I think you maybe point interpolate function)

def forward(self, input: Tensor) -> Tensor:
    return F.interpolate(input, self.size, self.scale_factor, self.mode, self.align_corners)