Hi,
I tried running the resnet based boundary attack, but I have a small question. According to the code the main loop should stop running, and finally it stops due to e_step > 500. When I checked the final adversarial image, it changed the image to eel, but the label for that image was not changed to sea lion (which is the aim of the attack). Can you explain why this issue is occurring?
Also, can you share whether your mean squared error value reduced to condition?
Hi,
I tried running the resnet based boundary attack, but I have a small question. According to the code the main loop should stop running, and finally it stops due to e_step > 500. When I checked the final adversarial image, it changed the image to eel, but the label for that image was not changed to sea lion (which is the aim of the attack). Can you explain why this issue is occurring?
Also, can you share whether your mean squared error value reduced to condition?