Team Name: A Cool Name
Group Members: Connor Thompson, Soroush Khadem, Greg Lund, Chris Beggs
To download all of the weights in the correct directory, from the main directory of this repo run the following
bash scripts/download_models.shTo download weights for an individual model, from the main directory run the following:
bash scripts/single_download.sh <model_name>model_name can be any of the following:
- pix2pix_dogs
- pix2pix_pizza
- pix2pix_trees
- pix2pix_apples
- cyclegan_pizza
- cyclegan_trees
- cyclegan_quickdraw_trees
- cyclegan_apples
December 7th: Final Presentation
December 9th: Final report due
Original paper: Image-to-Image Translation with Conditional Adversarial Networks
Tensorflow demo: https://www.tensorflow.org/tutorials/generative/pix2pix
Interactive demo: https://affinelayer.com/pixsrv/
UMich DL Course Slides: https://web.eecs.umich.edu/~justincj/slides/eecs498/498_FA2019_lecture20.pdf
GAN Tutorial(Goodfellow): NIPS 2016 Tutorial: Generative Adversarial Networks
Tips and Tricks to make GANs work: https://github.com/soumith/ganhacks
More stable GAN objective function: Wasserstein GAN and Improved training of Wasserstein GAN
Related latent space paper: Interpreting the Latent Space of GANs for Semantic Face Editing