search

OmidPoursaeed / Generative_Adversarial_Perturbations

Generative Adversarial Perturbations (CVPR 2018)
https://arxiv.org/abs/1712.02328
135 stars 22 forks source link

Do you use the whole training set of Imagenet? And could you share the parameters for imdep? #3

Closed theFool32 closed 5 years ago

theFool32 commented 5 years ago

Hi, thanks for your job and willing to share codes. I am working on similar project about imdep. However, you only give the example to train a universal one. Could you please share the parameters for imdep (e.g. batchSize, nEpochs)?

Thanks a lot.

OmidPoursaeed commented 5 years ago

Hi. We use mostly similar values as those for universal. You don't need to use the whole training set. Using a small subset of Imagenet suffice, and the model converges pretty fast.

theFool32 commented 5 years ago

Thanks for your response. :)