Need help with f-AnoGAN?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

tSchlegl
183 Stars 56 Forks MIT License 9 Commits 15 Opened issues

Description

Code for reproducing f-AnoGAN training and anomaly scoring

Services available

!
?

Need anything else?

Contributors list

No Data

f-AnoGAN: Fast unsupervised anomaly detection with generative adversarial networks

Overview

Code for reproducing f-AnoGAN training and anomaly scoring presented in "f-AnoGAN: Fast Unsupervised Anomaly Detection with Generative Adversarial Networks" (accepted manuscript). This work extends AnoGAN: "Unsupervised Anomaly Detection with Generative Adversarial Networks to Guide Marker Discovery".

Referencing and citing f-AnoGAN

If you use (parts of) this code in your work please refer to this citation:

Schlegl, T., Seeböck, P., Waldstein, S.M., Langs, G., Schmidt-Erfurth, U., 2019. f-AnoGAN: Fast Unsupervised Anomaly Detection with Generative Adversarial Networks. Medical Image Analysis 54, 30-44. DOI: https://doi.org/10.1016/j.media.2019.01.010

Prerequisites

  • Python (2.7), TensorFlow (1.2), NumPy, SciPy, Matplotlib
  • A recent NVIDIA GPU

f-AnoGAN building blocks

  • wgangp_64x64.py
    : Training a 64x64 WGAN architecture yields a trained generator (G) and discriminator (D). Modifies and extends (including
    tflib/
    ) Ishaan Gulrajani's Tensorflow implementation of the WGAN-GP model proposed in "Improved Training of Wasserstein GANs" (GitHub).
  • z_encoding_izif.py
    : Training the izi_f encoder (E) based on the trained WGAN model (G and D). Please refer to the full paper for more detailed information.
  • anomaly_detection.py
    : Code for anomaly scoring utilizing the trained G, D, and E.

Setting (image) paths

Image paths are set in

tflib/img_loader.py
. Images should be provided as "*.png" files structured in the following way: -
trainset_path
= "path-to-folder-holding-normal-training-images" -
trainset_val_path
= "path-to-folder-holding-normal-validation-images" -
test_normal_path
= "path-to-folder-holding-normal-test-images" -
test_anom_path
= "path-to-folder-holding-anom-test-images"

Please edit that file to specify the paths to your datasets.

Misc

Results of related research work are provided at CIR.

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.