Scoring the performance of GANs

This project is meant for evaluating GAN performances (Generative Adversarial Network).

Dependency

Make sure to run the project with Python 3.3 or greater. For those interested in Inception Scores and 1-NN Score with AlexNet, lateset version of tensorflow is need.

Features Supported

• Scoring GAN performances with these metrics:
  • Inception Score[to be completed]
  • 1-NN (One Nearset Neighbor) Score [basically done]
  • MAE (Mean Average Error) of pixel-wise color std and mean [basically done]
• Comparing two GANs against a common real distribution [basically done]
• Plotting the evolution of GAN performance as training proceeds [to be implementd]