Automatic image generation is a complex task with many applications in several domains such as security (e.g., gener- ating portraits from the description), styling and entertainment. In this project, advanced versions of Generative Adversarial Networks (GANs) are used to generate real images. A conditional GAN (cGAN) was implemented, followed by an evaluation with visual examination, k-nearest-Neighbours (kNN) and Fre ́chet Inception Distance (FID). The evaluation indicates that a cGAN can generate realistic images of handwritten digits. Whereas, the discussions shows that more work must be done to create realistic images from datasets which contains larger and more complex images.
You should know that.
Unfortunately, the data is not provided on a public host. So, please do it the old fashioned way and write an email, so that we can find a way.
conda env create -f environment.yml
conda activate mad40-env
You should know that as well :P
Just pull the repo, if you wanna change sth you can ask :)
Please do NOT commit any data files into the repositories. Data should always be kept seperate from code!
- Timo Bohnstedt - Coding, Report - GitHub Bohniti
Pretty much the BSD license, just don't repackage it and call it your own please! Also if you do make some changes, feel free to make a pull request and help make things more awesome!
The author would like to thank his Franz Koeferl and the MADI40-Team from the Machine Learning & Data Analytics Lab for excellent support.