ImportError: cannot import name 'dist_model'

[tianjiaozeng]

Thanks for sharing the pre-trained model. I'm very interested in this work but are facing a problem when running the code. Below is the error that occurs.

I'd really appreciate it if you could help me solve this problem. Many thanks!

ImportError Traceback (most recent call last)

in () 9 10 ---> 11 from utils import * 12 13 sys.path.append('models/') /content/drive/My Drive/LenslessLearning-master/utils.py in () 234 import sys 235 sys.path.append('/home/kristina/PerceptualSimilarity') --> 236 from models import dist_model as dm 237 from admm_helper_functions_torch import * 238 ImportError: cannot import name 'dist_model'

[antipa]

it looks like it's just having issues finding paths to various packages. The code is still looking for a specific path on someone else's computer--try modifying line 235 to point to whichever directory you installed PerceptualSimilarity and see if that helps? Maybe @monakhova could chime in and help out.

[monakhova]

The package is from the Perceptual Similarity repo and calls the LPIPS loss. You can find this here: https://github.com/richzhang/PerceptualSimilarity

[tianjiaozeng]

Thanks! It works now. Can I ask another question? I notice that in the admm_model.py, class ADMM_Net needs a parameter "h" for initialization. what should this "h" be if I'd like to retrain the models?

_class ADMM_Net(nn.Module): def init(self, batch_size, h, iterations, learning_options = {'learned_vars': []}, cuda_device = torch.device('cpu'), le_admm_s = False, denoisemodel = []):

Thank you!

The package is from the Perceptual Similarity repo and calls the LPIPS loss. You can find this here: https://github.com/richzhang/PerceptualSimilarity

[tianjiaozeng]

Thanks! It works now. Can I ask another question? I notice that in the admm_model.py, class ADMM_Net needs a parameter "h" for initialization. what should this "h" be if I'd like to retrain the models?

class ADMM_Net(nn.Module): def init(self, batch_size, h, iterations, learning_options = {'learned_vars': []}, cuda_device = torch.device('cpu'), le_admm_s = False, denoise_model = []):

Thank you!

[tianjiaozeng]

it looks like it's just having issues finding paths to various packages. The code is still looking for a specific path on someone else's computer--try modifying line 235 to point to whichever directory you installed PerceptualSimilarity and see if that helps? Maybe @monakhova could chime in and help out.

Can I ask another question? I notice that in the admm_model.py, class ADMM_Net needs a parameter "h" for initialization. what should this "h" be if I'd like to retrain the models?

class ADMM_Net(nn.Module): def init(self, batch_size, h, iterations, learning_options = {'learned_vars': []}, cuda_device = torch.device('cpu'), le_admm_s = False, denoise_model = []):

Thank you so much!

[monakhova]

h is the point spread function of the imaging system.

[tianjiaozeng]

h is the point spread function of the imaging system.

I notice in the DiffuserCam project, there's a mat file "example_psf.mat". Is this the same psf that should be used here?

Thank you so much!

[JayShaun]

The package is from the Perceptual Similarity repo and calls the LPIPS loss. You can find this here: https://github.com/richzhang/PerceptualSimilarity

Sorry, I don't know if I miss something, but I don't find 'dist_model' in directory 'models'. I'd really appreciate it if you could help me solve this problem.

[BaIGaO]

The package is from the Perceptual Similarity repo and calls the LPIPS loss. You can find this here: https://github.com/richzhang/PerceptualSimilarity

Sorry, I don't know if I miss something, but I don't find 'dist_model' in directory 'models'. I'd really appreciate it if you could help me solve this problem.

The latest repo of LPIPS has changed a lot. You can pull history repo ,here :https://github.com/richzhang/PerceptualSimilarity/releases/tag/1.0 .