search

xl-tang3 / UAUDeblur

(CVPR 23) Uncertainty-Aware Unsupervised Image Deblurring with Deep Residual Prior
57 stars 4 forks source link

Are any files missing? #1

Closed RHEPI closed 1 year ago

RHEPI commented 1 year ago

Is the w.moat file missing for the line w_mat = scio.loadmat(' w.moat ') in the UaUDeblur_demo.py file? If not, could you please provide the file for learning?

xl-tang3 commented 1 year ago

Sorry for the confusion. The w.mat file is the simulated true residual that was previously used to test and tune the model parameter. It's no longer necessary for the fixed model so I've deleted the relevant variable and statements in the updated version.

chunzizhu commented 1 year ago

Where is the supplementary material for this paper?

xl-tang3 commented 1 year ago

Where is the supplementary material for this paper?

Updated in the repository.

chunzizhu commented 1 year ago

Is UaUDeblur_demo.py code used for training? Would it be convenient to publish the code for the test?

ironman0086 commented 1 year ago

In debugging the file "UaUDeblur_demo.py", there is a error: self.img_est_np = torch_to_np(self.img_est_torch).squeeze() AttributeError: 'DeepResPrior' object has no attribute 'img_est_torch' How could I revise this error?

xl-tang3 commented 1 year ago

line 140 'self.img_est_torch = torch.real(torch.fft.irfft2(torch.fft.rfft2(self.img_out_torch) * self.PSF))'

Did you have declared a new variable 'img_est_torch' or comment out the line 140? There's literally no attribute 'img_est_torch' in the latest version.

Marsbegi commented 1 year ago

Hello, when I run UaUDeblur_demo.py according to the readme file, there is a problem of inconsistent dimensions. I found that the dimensions of the variables self.img_est_torch and self.img_blurred_torch are inconsistent when calculating mse. How could I revise this error

ironman0086 commented 1 year ago

李付谦已收到。

xl-tang3 commented 1 year ago

Hello, when I run UaUDeblur_demo.py according to the readme file, there is a problem of inconsistent dimensions. I found that the dimensions of the variables self.img_est_torch and self.img_blurred_torch are inconsistent when calculating mse. How could I revise this error

It’s been noticed that many attempts have ended up with the same issue. To the best of my knowledge, it is concerned with the FFT transformation on line 139 since the experiment is implemented on PyTorch 1.10.1, which is an old version. Therefore, it’s recommended to try another version or find alternative functions for torch.fft.xxx.

Marsbegi commented 1 year ago

I get it.Thanks for your reply.

---Original--- From: @.> Date: Sat, Jul 29, 2023 23:26 PM To: @.>; Cc: @.**@.>; Subject: Re: [xl-tang01/UAUDeblur] Are any files missing? (Issue #1)

Hello, when I run UaUDeblur_demo.py according to the readme file, there is a problem of inconsistent dimensions. I found that the dimensions of the variables self.img_est_torch and self.img_blurred_torch are inconsistent when calculating mse. How could I revise this error

Hello, when I run UaUDeblur_demo.py according to the readme file, there is a problem of inconsistent dimensions. I found that the dimensions of the variables self.img_est_torch and self.img_blurred_torch are inconsistent when calculating mse. How could I revise this error

It’s been noticed that many attempts have ended up with the same issue. To the best of my knowledge, it is concerned with the FFT transformation on line 139 since the experiment is implemented on PyTorch 1.10.1, which is an old version.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

ironman0086 commented 1 year ago

李付谦已收到。