search

tum-vision / learn_prox_ops

Implementation of "Learning Proximal Operators: Using Denoising Networks for Regularizing Inverse Imaging Problems"
https://arxiv.org/abs/1704.03488
GNU General Public License v3.0
44 stars 14 forks source link

module 'proximal' has no attribute 'patch_BM3D' #1

Closed Benlawei closed 6 years ago

Benlawei commented 6 years ago

Hello,I meet some issues,I use Ubuntu16.04 LTS and my gpu is GeForce GTX 1060 when I use the command "python3 src/experiment_demosaicking.py grid_search_all_images with elemental.denoising_prior=BM3D",I got a Traceback "AttributeError: module 'proximal' has no attribute 'patch_BM3D'",but I can find "patch_NLM" in the path "proximal/prox_fns/patch_NLM.py",Is it missing a "patch_BM3D.py" there?

timmeinhardt commented 6 years ago

Hello and thank you for your interest in our research. In order to use our code you should follow the installation steps in our README.rst. Your error comes from using the official ProxImaL package instead of our own modified version. Please install our version with pip3 install git+https://github.com/timmeinhardt/ProxImaL@learn_prox_ops. This includes the BM3D denoising proximal function.

If this resolves your issue please close it and do not hesitate to ask if you encounter further problems!

Benlawei commented 6 years ago

Thank u,I have followed the installation steps and solved above problem. But when I use python3 src/experiment_deblurring.py with experiment_name=experiment_a image_name=barbara elemental.optimal_DNCNN_experiment_a,I got a lot of Traceback,such as: INFO:tensorflow:Restoring parameters from models/DNCNNgaussian_0.0240-40-1128/model.ckpt INFO - tensorflow - Restoring parameters from models/DNCNN__gaussian_0.0240-40-1__128/model.ckpt 2018-04-06 20:21:00.430443: W tensorflow/core/framework/op_kernel.cc:1192] Data loss: not an sstable (bad magic number) ... DataLossError (see above for traceback): not an sstable (bad magic number) [[Node: save/RestoreV2_5 = RestoreV2[dtypes=[DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save/Const_0_0, save/RestoreV2_5/tensor_names, save/RestoreV2_5/shape_and_slices)]] [[Node: save/RestoreV2_14/_5 = _Recv[client_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_76_save/RestoreV2_14", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]()]] I think there may be some problems when restoring parameters from checkpoint.

timmeinhardt commented 6 years ago

Could you please check if your TensorFlow version is indeed 1.3.0.

Benlawei commented 6 years ago

I used the command pip3 install tensorflow-gpu==1.3.0,and my TensorFlow version is tensorflow-gpu (1.3.0). In fact, I have used other versions of the TensorFlow as an attempt，but I still got the same TraceBack .

timmeinhardt commented 6 years ago

And you also installed Git LSF and downloaded all the necessary model files? There must be three files in the corresponding subdirectory.

Edit.: I just reproduced the error when I clone the repository without Git LSF. Eventhough there are files in the model directories these are just 4kB placeholders.

timmeinhardt commented 6 years ago

@Benlawei Can we close this issue or did you encounter any other troubles?