Playing with Wgans and Gans + Flexible Checkpoint Saver for Pytorch

Dataset - Color FERET

FERET database of facial images collected under the FERET program, sponsored by the DOD Counterdrug Technology Development Program Office

Main Objectives

• Understand gans and wgans.
• Obtain and create Dataloader for images of faces.
• Create wgan for creating faces.
• Create flexible Checkpoint Saver for pytorch.
• Compare dense layers with deconvolution layers.
• Create wgan for text.
• Visualize training f.ex loss, compare with other types of models.
• Compare Conditional improved WGAN which replace cliping gradient with penalty.

Requirements

• Python >= 3.6
• Pytorch >= 0.4
• Color Feret Dataset