Skip to content

Question about the shape and data dtype of lowpass ? #59

Description

@FeiSong123

I ran the code and got the outputs below:

import torch
from pytorch_wavelets import DTCWTForward, DTCWTInverse, ScatLayer
xfm = DTCWTForward(J=3, biort='near_sym_b', qshift='qshift_b', include_scale=True)
X = torch.randn(2,3,512,512)
Yl, Yh = xfm(X)
for l in Yl:
    print(l.shape, l.dtype)
print(Yh[0].shape, Yh[0].dtype)
print(Yh[1].shape, Yh[1].dtype)
print(Yh[2].shape, Yh[2].dtype)
torch.Size([2, 3, 512, 512]) torch.float32
torch.Size([2, 3, 256, 256]) torch.float32
torch.Size([2, 3, 128, 128]) torch.float32
torch.Size([2, 3, 6, 256, 256, 2]) torch.float32
torch.Size([2, 3, 6, 128, 128, 2]) torch.float32
torch.Size([2, 3, 6, 64, 64, 2]) torch.float32

And I am confused about Why is the Resolution of Lowpass Filters Double That of Highpass Filters at the Same Level, and Why Isn't the Data Type of Lowpass Complex?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions