colinskot/generate-faces

GAN generates new celebrity faces

Generate Faces

A Generative Adverserial Network (GAN) produces fake celebrity faces. From a dataset of over 200,000 low resolution pictures that are cropped to 28x28, this machine learning architecture learns to generate images and discriminate it's own creation.

Getting Started

Simply run the Jupyter Notebook dlnd_face_generation.ipynb or you can run the script face_generation.py

python face_generation.py

I Just Want to See the Results!

Just open dlnd_face_generation.ipynb here on Github and you can view the results of the project!

Prerequisites

You can install the required packages through Anaconda's environment manager using the machine-learning.yml file

conda env create -f machine-learning.yml

Then, activate the environment and run face_generation.py

activate machine-learning

Otherwise, check out the machine-learning.yml file for dependencies and their versions

Running the tests

Simply add test cases to problem_unittests.py or run it

python problem_unittests.py

Built With

• TensorFlow - The machine learning framework
• Anaconda - The environment manager
• Jupyter Notebook - The code documentation