Image of Research
 

Guard Against Adversarial Attack

Description

You can fool Deep Neural Network really easily, even by changing just one pixel. My research interest lies in making robust model in the computer vision domain. This picture shows, in the right part self-driving car is able to recognize the stop sign but in the left, it recognizes it as speed limit 45. The left part of the image was intentionally perturbed which fools the model of self driving car.