defensegan icon indicating copy to clipboard operation
defensegan copied to clipboard

Published 20 hours ago • verified publisher icon kabkabm
→

Metadata

Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models (published in ICLR2018)

Results 14 defensegan issues
Sort by recently updated
recently updated
newest added

Update requirements.txt

Minor change to the requirements.txt file

AtishayJain-AJ avatar AtishayJain-AJ

WGAN-GP loss function : Equation is not same as presented in WGAN-GP paper

In Model/gan.py line **#177** and **#178** : differences = self.fake_data - self.real_data interpolates = self.real_data + (alpha * differences) According to my understanding of WGAN-GP paper interpolates should be interpolates...

shubham-malaviya avatar shubham-malaviya

Test on a single Already Perturbed Image

2 comment

Is there any way to use the already trained celeba model (trained with train.py) and infer for a single perturbed image and see if the model regenerates the original?

imransalam avatar imransalam

Here is there a problem with the url link of the celebA dataset and it cannot be accessed

xckBUPT avatar xckBUPT

← Metadata

224
Stars
63
Forks
Watchers

Owner

verified publisher icon kabkabm

Metadata

Defense-GAN: Protecting Classifiers Against Adversarial Attacks Using Generative Models (published in ICLR2018)

Back