Adversarial.io currently creates adversarial images against one single image recognition model: Google Inception V.3.
It works best with 299 x 299px images that depict one specific object. Examples below.
As this project progresses, more image recognition models and adversarial techniques will be added. Learn more…
* Sometimes even the original image is not recognized correctly, which is based in the failure rate of Inception V.3 of 22%. More recent versions show lower original failure rates.