ustclby / unsupervised-domain-specific-deblurring Goto Github PK

Thank you for your work. Is it on Windows or Ubuntus? Which version of PyTorch?

Hi, thanks for your code. I want to know how long does it take to train the model?

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [80, 3, 1, 1]] is at version 2; expected version 1 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True).
python-BaseException

Hello
How are you?
Thanks for contributing to this project.
I am going to use this method to a general image restoration (denoise, deraining, etc).
Do u think it is possible?

Hi,
Thanks for your code. But I met some problem when I run this code

/pytorch/aten/src/THCUNN/BCECriterion.cu:42: Acctype bce_functor<Dtype, Acctype>::operator()(Tuple) [with Tuple = thrust::detail::tuple_of_iterator_references<thrust::device_reference, thrust::device_reference, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type>, Dtype = float, Acctype = float]: block: [0,0,0], thread: [30,0,0] Assertion input >= 0. && input <= 1. failed.
/pytorch/aten/src/THCUNN/BCECriterion.cu:42: Acctype bce_functor<Dtype, Acctype>::operator()(Tuple) [with Tuple = thrust::detail::tuple_of_iterator_references<thrust::device_reference, thrust::device_reference, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type, thrust::null_type>, Dtype = float, Acctype = float]: block: [0,0,0], thread: [31,0,0] Assertion input >= 0. && input <= 1. failed.
Traceback (most recent call last):
File "train.py", line 87, in
main()
File "train.py", line 58, in main
model.update_D(images_a, images_b)
File "/home/dengzeyu/run_on_server/Unsupervised-Domain-Specific-Deblurring-master/src/model.py", line 191, in update_D
loss_D1_A = self.backward_D(self.disA, self.real_I_encoded, self.fake_I_encoded)
File "/home/dengzeyu/run_on_server/Unsupervised-Domain-Specific-Deblurring-master/src/model.py", line 222, in backward_D
ad_fake_loss = nn.functional.binary_cross_entropy(out_fake, all0)
File "/home/dengzeyu/anaconda3/envs/py36/lib/python3.6/site-packages/torch/nn/functional.py", line 2113, in binary_cross_entropy
input, target, weight, reduction_enum)
RuntimeError: reduce failed to synchronize: device-side assert triggered
Do you know the reason？
Best

Hi, I had a question while reading your paper. In the section on the data set in your paper, you mentioned the following: "We randomly split the whole dataset into three mutually exclusive subsets: sharp training set (100K images), blurred training set (100K images) and test set (2137 images). For the blurred training set, we use the method in Section 3.6 to blur the images." But we did not find out exactly what test set you used in this sentence. I would like to know exactly what test set was used to evaluate performance.

i want to apply it on image-denosing， but it does no sense... There r many artifacts...

i have read your code carefully, there are two confusions in my head:

1. first, why add guassian noise in Content Encoder part
2. second,why exists four discriminators: disB2 and disA2？ is it indispensable? can i cancel it?

Would you provide the trained model?

Do you use spectral norm in the discriminators?
Thank you.

Dear author,

In the DRIT model you have taken inspiration from, gradient clipping is applied to stabilize gan training. I couldn't see that in your implementation if I am not mistaken. I was wondering if this is intentional.

Thank you.

this is my code for testing :python test.py --dataroot ../datasets/test/test_blur/ --num 1 --resume ../models/VGGFace16.pth --name job_name --orig_dir ../datasets/test/test_orig --percep face --result_dir ../datasets/test/results
And error occured:
Traceback (most recent call last):
File "test.py", line 105, in
main()
File "test.py", line 41, in main
model.resume(opts.resume, train=False)
File "D:\Study\deblur\Unsupervised-Domain-Specific-Deblurring-master\src\model.py", line 326, in resume
self.enc_c.load_state_dict(checkpoint['enc_c'])
KeyError: 'enc_c'
There is no result, how to solve the problem?

Hi,
Thanks for your code. I wonder if the code can be adapted to support multi-gpus?

Best,

Hi, after reading your paper, I found that you used three datasets to train the model. But I am not sure whether you use each dataset to train the model individually or put all data together to train the model?If possible, can you provide the pre-processed relevant data?