Introduction

In this project, we compared the traditional CDCGAN (Conditional Deep Convolutional Generative adversarial network) with PCDCGAN (Predictive Deep Convolutional Generative Adversarial Network)

How to start

First, clone the repository to your computer, then open the command line on the path of the cloned repository.

To use pretrained models

1. Download one of the models from the following link: https://drive.google.com/drive/folders/1PWxUckhEZdz7VetuqyQUffeUO1SWzh4i?usp=sharing
   This files contains the pretrained models of both architectures. After you've downloaded the file place it inside the root of the repository folder.

2. To create images from the models type in the command line:
   python create.py "file_name_without_pkl" for example python create.py "pcdcgan_alpha_001"

3. To evaluate the model with Inception Score and Frechet Inception Distance type in the command line:
   python evaluate.py "file_name_without_pkl"

Train and use a new models

To train and use a new pcdcgan model type in the command line:
python pcdcgan.py "file_name_without_pkl"

To train and use a new cdcgan model type in the command line:
python cdcgan.py "file_name_without_pkl"

You can keep training the new models or pretrained models with the same command.
Enjoy!