The Jupyter Notebook in this repository(GAN_Tutorial.ipynb) provides a step-by-step guide on how to build, train and evaluate a GAN using Tensorflow. The notebook covers the following:
- Data Preprocessing: How to load in, normalize, and batch your data for training
- Model architecture: How to create the generator and discriminator networks
- Training loop: How to train the models on a batch of data, and how to implement a training loop
- Evaluation: How to generate new images using our trained GAN
- Saving model for later use
The requirements needed for this project are listed in the requirements.txt file.
To get started clone the repository:
git clone https://github.com/shreyvish5678/Generative_Adversarial_Network.git
Then install the requirements:
pip install -r requirements.txt
After that, install Jupyter Notebook:
pip install notebook
Then, you can open up the notebook with:
jupyter notebook GAN_Tutorial.ipynb
The pre-trained models are also in the repository included as generator.h5 and discriminator.h5
There is also more information included on GANs in GANS.pdf
You can test the gan in test_gan.ipynb