rebeccaeexu / dtnet Goto Github PK

Thank you very much for your work, I'm having some problems with your code parts.About this codeDD[:, :, i] = torch.fft.ifft(im[:, :, i], norm='ortho', dim=-1).real
What version of torch are you using? My version is 1.6 and it doesn't seem to have torch.fft.ifft, I looked up some information and the old version seems to be this codetorch.irfft(output_fft_old , signal_ndim=2, normalized=False, onesided=False)
However, in the new version, the input data type is complex, and in the old version, it is a two-dimensional vector formed by the real part and the imaginary part, but the Fourier transformation process of the feature in your code does not seem to usetorch.fft.fft，I'm not quite sure what type of dataim[:, :, i] is.Do you have a way to invert theim[:, :, i]without upgrading the torch version?

Why you put alignment after demoieing？ I've found that there is some instability when using the PCD alignment in the VDmoire paper.Is it based on this aspect of your consideration that you put demoireing before alignment?