agonzgarc / cross-domain-disen Goto Github PK

Hi @agonzgarc , I'm very interested in your work and have read your paper carefully. In Section 2.1, you mentioned that directly calculating the L1 loss between the shared codes of two different domains may not increase the similarity between them, so you proposed to add small noise to the output of the encoder to yield better results. However, I found that you did not add such noise in your code, may I ask why? Looking forward to your reply!

hi @agonzgarc , I was very interested in your work and read your code carefully.In the "ops. py" file, I don't quite understand the role of "preprocess" and "deprocess". I hope you can help me

Hi,

I am curious about how to evaluate cross-domain retrieval in order to reproduce Table 1 in the paper? Can you provide codes or references for it, please?

Any help is appreciated!

@agonzgarc Awsome work! I am very inspaired by your work!
And may I ask how many epochs this code needs to run? I found that the code could not stop, and I can not find where to define the epochs. Forgive that I am a new learner about tensorflow...

and Whether there is code for retrieve?

Please excuse me, I meet the error "No module named maxpool" at line 16 of "ops.py" when running the code "run_cross_domain_disen.py". I wonder if there are any missing documents, for instance, "maxpool.py".

Hi, Im trying to download MNIST-CDCB dataset.
But the download link is invalid now,
Can I get another way to get the dataset?

Thank you

Hi, I used the pretrained model you provided to replicate the MNIST_CDCB results but the model produces noisy image as shown in the image below. I am using the latest version of the code without modifications. Any ideas on what the issue could be?
Thank you

Hi @agonzgarc ,I have read your code carefully, but I have not found the discriminator GAN module. Did I miss something?
Looking forward to your reply!