Skip to content

Generative Adversarial Stress Test Networks

License

Notifications You must be signed in to change notification settings

LuukdeJong123/gasten

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GASTeN Project

License

Variation of GANs that, given a model, generates realistic data that is classified with low confidence by a given classifier. Results show that the approach is able to generate images that are closer to the frontier when compared to the original ones, but still realistic. Manual inspection confirms that some of those images are confusing even for humans.

Paper: GASTeN: Generative Adversarial Stress Test Networks

Create Virtual Environment and directories

mamba create -n gasten python=3.10

mamba activate gasten

mamba install pip-tools

pip3 install -r requirements.txt

mkdir <file-directory>/data/clustering

mkdir <file-directory>/data/fid-stats

mkdir <file-directory>/out

Run

env file

Create .env file with the following information

CUDA_VISIBLE_DEVICES=0
FILESDIR=<file-directory>
ENTITY=<wandb entity to track experiments>

GASTeN

Run AutoGASTeN to create images in the bounday between 8 and 9.

python3 -m src.optimization --dataset mnist --pos 8 --neg 9 --config experiments/optimization/auto_gasten.yml --config_clustering experiments/clustering/mnist_7v1.yml

About

Generative Adversarial Stress Test Networks

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%