fastgradient

USER MANUAL

To run this adversarial attack requires a GPU server. In order to be able to run one needs to follow the following steps:

Setup the environment.

Install pytorch on the gpu server

pip install torch==1.6.0 torchvision==0.7.0

Clone repository from github.

git clone https://github.com/takchiks/fastgradient

Download dataset from:

https://shapenet.cs.stanford.edu/media/modelnet40_normal_resampled.zip

Extract zip and copy folder modelnet40_normal_resampled to data folder

python adversary.py

and generate adversary examples

To visualize

cd visualizer

python show3d_balls.py

python show3d_balls_perturb.py

python show3d_balls_test.py