Open barghouthi opened 4 years ago
It is similar to GAN but the task is not "generative". The attacker here tries to mask the most important region of images away so as to fool the prediction of the discriminator.
Sure. I will update the report for this week soon.
By attacker/discrim, are you doing a GAN-like training? Let me know how progress is going and if you have any qs.