Excuse the link direct to a PDF but I can't find an html version. Interesting paper about hacking machine vision systems to "see" what you want them to see, including examples. Abstract: "We...
Excuse the link direct to a PDF but I can't find an html version. Interesting paper about hacking machine vision systems to "see" what you want them to see, including examples.
Abstract: "We present a method to create universal, robust, targeted adversarial image patches
in the real world. The patches are universal because they can be used to attack
any scene, robust because they work under a wide variety of transformations,
and targeted because they can cause a classifier to output any target class."
I think, when lacking a proper HTML version, it would be courteous to link it's actual place of archive as well. If only so people can easily bookmark it or add it to their bibliographies.
I think, when lacking a proper HTML version, it would be courteous to link it's actual place of archive as well. If only so people can easily bookmark it or add it to their bibliographies.
Excuse the link direct to a PDF but I can't find an html version. Interesting paper about hacking machine vision systems to "see" what you want them to see, including examples.
Abstract: "We present a method to create universal, robust, targeted adversarial image patches
in the real world. The patches are universal because they can be used to attack
any scene, robust because they work under a wide variety of transformations,
and targeted because they can cause a classifier to output any target class."
I think, when lacking a proper HTML version, it would be courteous to link it's actual place of archive as well. If only so people can easily bookmark it or add it to their bibliographies.
Good point, my apologies for not thinking of that.