search

MadryLab / cifar10_challenge

A challenge to explore adversarial robustness of neural networks on CIFAR10.
MIT License
488 stars 133 forks source link

Image Channels #11

Closed rajasekharponakala closed 5 years ago

rajasekharponakala commented 5 years ago

Hi Team,

For example, if a CNN model (say, image classification) is trained on a 1-channel (grayscale) inputs, how can we deal with the perturbations or l-norm constraints? any thoughts? Thanks.

dtsip commented 5 years ago

Hi, perturbations are defined based on the single channel inputs. An L-infinity perturbations allows the adversary to perturb each pixel of this grayscale image by epsilon.