/gtsrb-adversarial

Primary LanguageJupyter Notebook

Adversarial Attacks on Traffic Sign Recognition System

This was the source code for my masters thesis.

Go to https://kimjip.github.io/gtsrb-adversarial/ for a quick explanation and sample images.

• Training a CNN with clean images, then testing it with traffic sign images that are slightly modified but unnoticeable to humans

• Using Explainable AI (Grad-CAM) to explain image misclassifications