Offical implemention of Robust Superpixel-Guided Attentional Adversarial Attack (CVPR2020)

Jan 06, 2022 1 min read

Super_Pix_Adv

Setup

Install python — This repo is tested with 3.6
Install PyTorch version >= 1.0.0, torchvision >= 0.2.1
Download Inception-v3 pretrained model from https://download.pytorch.org/models/inception_v3_google-1a9a5a14.pth and move it to ./PRETRAIN/

Super-Pixel Generation

python -u get_slic.py

K: number of super pixels

M: compactness factor

We resize the image to 299×299 as pre-process and 400 images example can be found at here

Adversarial Attack

run on ImageNet dataset for white-box attack succes rate and robustness toward resize(2x)

python -u abim_sup.py

GitHub

View Github

Robust Attack

John was the first writer to have joined pythonawesome.com. He has since then inculcated very effective writing and reviewing culture at pythonawesome which rivals have found impossible to imitate.