vl_simplenn_tidy.m is missing about pnp-sci HOT 5 CLOSED

I think MatConvNet is missing here, since it is required to apply FFDNet as a denoser. Please refer to the doc to compile it (with CUDA and CuDNN for better performance on GPU). You might find this issue useful if you compile it on MATLAB 2018a and above.

I will try to release the real data and Python (PyTorch) version hopefully after we complete the journal version.

from pnp-sci.

Problem resolved. MatConvNet was missing. Thank you very much!

from pnp-sci.

May I ask if we need to run on GPU in order to reproduce some results? -- Thanks!

from pnp-sci.

GPU is not needed to reproduce the same results in terms of reconstruction quality. You can set para.useGPU = false; in line#82 of test_pnpsci.mto use your local CPU for running the pre-trained network. However, MatConvNet is required for running the MATLAB version of FFDNet. Sadly, MatConvNet only supports up to CUDA 10 and is no longer maintained. You may check the Python version (PyTorch is used) works on CPU only as well (Ooops, I need to update the readme of PnP-SCI_python). Again, you need to set useGPU = False in line#183 of pnp_sci_video_grayscale_benchmark.py or multiple lines of pnp_sci_video_middle_scale_color.py.

from pnp-sci.

@liuyang12 Thank you! super clear. I have already starred and cloned your python version. It can be a great help to my course project. Thank you!

from pnp-sci.