Final project for DD2424 Deep Learning in Data Science at KTH Royal Institute of Technology, Stockholm.
Create fake human faces with DCGANs. Evaluation performed using Fréchet Inception Distance (FID).
- Diogo Pinheiro (https://github.com/DiogorPinheiro)
- Jakob Lindén (https://github.com/jakobGTO)
- Márk Csizmadia (https://github.com/mark-antal-csizmadia)
- Patrick Jonsson (https://github.com/hiver-py)
Download the data sets either from original source or Kaggle, then extract the images (cropped and aligned) to data/celebA/.
- Learning Rate: 1e-4
- Beta1 : 0.5
- Batch Size : 128
- Leaky ReLU in both discriminator and generator
- TensorFlow
- Google Collab
[1] Alec Radford, Luke Metz, and Soumith Chintala. Unsupervised representation learning withdeep convolutional generative adversarial networks, 2016