when I run the test.py on pf-pascal dataset, it error with FileNotFoundError: [Errno 2] No such file or directory: '/home/xx/chm/Datasets/PF-PASCAL/test_pairs.csv'. The pf-pascal dataset I used is downloaded from https://www.di.ens.fr/willow/research/proposalflow/. The url of pf-pascal dataset in download.py cannot connected, https://docs.google.com/uc?export=download?id=1OOwpGzJnTsFXYh-YffMQ9XKM_Kl_zdzg return 400.

Created conda environment with python=3.7, I tried all available versions of pytorch=1.3.1 (paired with fixed torchvision=0.4.2 version) with different cudatoolkit versions (9.2, 10.0, 10.1 ) it seems like there's some problem between torchvision0.4.2 and PIL, causing a failure in importing torchvision properly.

Then I tried

conda install pytorch=1.4.1 torchvision cudatoolkit=10.1 -c pytorch

it worked and but the results I got are quite different.

I wonder what pytorch versions and cudatoolkit versions are recommended.

Hello! when I run the test.py on pf-willow dataset, it error with FileNotFoundError: [Errno 2] No such file or directory: '/home/xx/chm/Datasets/PF-WILLOW/test_pairs.csv'. The PF-WILLOW dataset I used is downloaded from https://www.di.ens.fr/willow/research/proposalflow/. Please help me ! Please!

Hello @juhongm999 !

Thank you for such great work.

I am experiencing an issue with the new weights located on your google drive.

If I try to use them with pytorch 1.3.1 I get the following error:

ValueError                                Traceback (most recent call last)
~/.anaconda3/envs/chm/lib/python3.7/tarfile.py in nti(s)
    186             s = nts(s, "ascii", "strict")
--> 187             n = int(s.strip() or "0", 8)
    188         except ValueError:

ValueError: invalid literal for int() with base 8: '_tensor_'
 ...
~/.anaconda3/envs/chm/lib/python3.7/site-packages/torch/serialization.py in _load(f, map_location, pickle_module, **pickle_load_args)
    597             if _is_zipfile(f):
    598                 # .zip is used for torch.jit.save and will throw an un-pickling error here
--> 599                 raise RuntimeError("{} is a zip archive (did you mean to use torch.jit.load()?)".format(f.name))
    600             # if not a tarfile, reset file offset and proceed
    601             f.seek(0)

RuntimeError: pas_psi.pt is a zip archive (did you mean to use torch.jit.load()?)

The full error could be found here.

But if I use the 1.8.1 version of pytorch I get the following error.

  Missing key(s) in state_dict: "learner.conv2ds.0.weight", "learner.conv2ds.1.weight", "learner.conv2ds.2.weight", "learner.chm6d.bias", "learner.chm6d.param". 
	  Unexpected key(s) in state_dict: "learner.conv2ds.0.0.weight", "learner.conv2ds.0.1.weight", "learner.conv2ds.0.2.weight", "learner.chm6d.0.bias", "learner.chm6d.0.param.0", "learner.chm6d.0.param.1", "learner.chm6d.0.param.2", "learner.chm6d.0.param.3". 
	  size mismatch for learner.chm4d.weight: copying a param with shape torch.Size([55]) from checkpoint, the shape in current model is torch.Size([1, 1, 5, 5, 5, 5]).

I believe this happens because the weights are for a slightly newer network, then the one currently in the repo. Is there any chance you could update the network in the repository ?

Thanks in advance, Arseny

Hi @juhongm999, Thanks for your great work.

When I run the train.py on pfpascal and spair, respectively, it errors with OSError: Not a gzipped file (b'<!') and tarfile.ReadError: not a gzip file. I think the error occurs line 44 in download.py.

Could you check it?

the command I did : python train.py --ktype psi --benchmark spair

Hi @juhongm999 ,

Thanks for your great work. I want to reproduce your results and visualization. How can we draw images of correspondence between 2 images? (e.g. cat images in Figure 1 of your paper). Do you have any implementation for that? Thanks!!

Best, Hai