https://github.com/airaria/adversarialattackinceptionv3
Attack InceptionV3 net using FGM( fast gradient method) and show saliency maps.
https://github.com/airaria/adversarialattackinceptionv3
attack deep-learning inceptionv3 tensorflow
Last synced: 5 months ago
JSON representation
Attack InceptionV3 net using FGM( fast gradient method) and show saliency maps.
- Host: GitHub
- URL: https://github.com/airaria/adversarialattackinceptionv3
- Owner: airaria
- Created: 2017-09-03T14:36:19.000Z (about 8 years ago)
- Default Branch: master
- Last Pushed: 2017-11-09T13:46:53.000Z (almost 8 years ago)
- Last Synced: 2025-04-05T06:31:45.816Z (6 months ago)
- Topics: attack, deep-learning, inceptionv3, tensorflow
- Language: Python
- Size: 99.6 KB
- Stars: 13
- Watchers: 2
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Simple adversarial attack on InceptionV3
Attack InceptionV3 net using FGM( fast gradient method) and show saliency maps.
## Prerequisites
- Python 3.5+
- tensorflow 1.2+
- sklearn, numpy, pandas, skimage
## Dataset and model
Download the dataset and model file from Kaggle: https://www.kaggle.com/c/nips-2017-non-targeted-adversarial-attack/data
+ Download Inception-v3 model ckpt file and put it in ./ .
+ Download the 1000-images development-set zip file and unzip the *images* directory to ./images, such that the image files are located in ./images/images/.
+ Finally, If you would like to generate and save the saliency maps, uncomment the lines following
```python
# generate saliency maps
```
in main.py.
## Acknowledgements
Inspired by Kaggle competitions on adversarial attacks and tensorflow/cleverhans.