Abstract: Adversarial examples are commonly viewed as a threat to ConvNets. Here we present an opposite perspective: adversarial examples can be used to improve image recognition models if harnessed ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results