The research project was associated with "[INF-DSAM9] Computational Foundations of Data Science: Deep Learning", Summer Semester 2020, for my Masters of Science: Data Science, University of Potsdam, Germany.
You can find the Technical Report on ResearchGate.
Train a conditional generative adversial network on the CelebA dataset. The result should be a model that generates instances of faces with the desired attribute.
The notebook is compatible with Google Colab.
How to run the notebook:
- Download the notebook "N3_GAN_CelebA_final.ipynb" along with all other "*.py" files into your Colab working directory (indicated by the folder icon on the left when you have a colab notebook open).
- Change the path directories in "default" argument of parser.add_argument() function according to your need.
- The first part of the notebook is for data generation. Restart the runtime after executing this part.
- Load the training in the second part. Let it train.
There is also a version that can be used locally (N3_GAN_local.ipynb) that was used in development and might be easier to understand for some.
See the Colab Demo for a small demonstration of results.