dkk / DCGAN256

A general purpose DCGAN that generates 256 x 256 RGB images. Tested with the 11K hands Dataset

DCGAN256

A DCGAN used to train a hand generator that outputs 256 x 256 RGB images using the 11K hands Dataset.

For more details, see the jupyter notebook.

Environment

The jupyter notebook was implemented to run in Google Colab, but it would be easy to change it to use a local or remote instance instead.

Result

The following GIF shows the output of the model in about the first 1000 epochs:

References

Losely based on

1. DCGAN256 by t0nberryking
2. DCGAN256 by manicman1999
3. DCGAN image generator by gsurma
4. Tensorflow DCGAN
5. Tensorflow Image Preprocessing
6. Dataset: Mahmoud Afifi, "11K Hands: Gender recognition and biometric identification using a large dataset of hand images." Multimedia Tools and Applications, 2019.
7. Medium: 10 Lessons I Learned Training GANs for one Year
8. How to Train a GAN? Tips and tricks to make GANs work
9. Medium - GAN — Ways to improve GAN performance
10. Medium - Generating High-Resolution Images Using Deep Autoregressive Models