LazarValkov / GanModeCollapseEvaluation

Code for training Generative Adversarial Networks (GANs) and evaluating the models' mode collapse

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Mode Collapse Evaluation in GANs

Code for training Generative Adversarial Networks (GANs), and evaluating the models' mode collapse. The code was taken and adapted from https://github.com/carpedm20/DCGAN-tensorflow .

The following models are supported:
Evaluating mode collapse using the following 2 methods, introduced in https://arxiv.org/pdf/1611.02163.pdf :
  • Number of modes covered with the Stacked Mnist dataset (section 3.3.1, Discrete Mode Collapse)
  • Inference via Optimisation (section 3.4.1)

Requirements:

  • Tensorflow
  • SciPy
  • Keras (only used to import keras.optimizers.Adam )

Datasets

Running

The Run_*.py files contain preset configurations for each model/dataset.

About

Code for training Generative Adversarial Networks (GANs) and evaluating the models' mode collapse

License:MIT License


Languages

Language:Python 100.0%