Humans can decipher adversarial images

HIGHLIGHTS

• who: Zhenglong Zhou from the Hopkins University have published the research: Humans can decipher adversarial images, in the Journal: NATURE COMMUNICATIONS NATURE COMMUNICATIONS
• what: A small minority of the images in the present experiments (e_g, 3/48 in Experiment 1) had CNN-generated labels that were actively rejected by human subjects, who failed to pick the CNN's chosen label even compared to a random label drawn from the imageset. The conclusions of this work are consonant with the own , in that they show how humans and CNNs can be made to give similar | https . . .
  

   

  

  If you want to have access to all the content you need to log in!

  Thanks :)

  Username or Email

  Password

  Remember me

  Lost your password?

  If you don't have an account, you can create one here.