I have two questions about this function in line66 of SinGAN/imresize.py.
def imresize_in(im, scale_factor=None, output_shape=None, kernel=None, antialiasing=True, kernel_shift_flag=False):
Why did you chose "cubic" as default kernel method?
Why did you do resizing dim by dim, instead of using simple method such as "nn.Upsample with scale_factor<1" ?
I have two questions about this function in line66 of SinGAN/imresize.py.
def imresize_in(im, scale_factor=None, output_shape=None, kernel=None, antialiasing=True, kernel_shift_flag=False):