ango-LookingOutwards02
Mario Klingemann
I’m training a next-frame-prediction #pix2pix model on some fireworks footage. pic.twitter.com/o0vd7njAai
— Mario Klingemann (@quasimondo) January 2, 2018
Feedback loops never cease to amaze me. Starting with the first seed image the rest of the clip is generated by feeding the output back into the input of the GAN. pic.twitter.com/gX4qN5oTv7
— Mario Klingemann (@quasimondo) January 3, 2018
This system is pretty unstable. After every epoch of training it behaves differently. Quite a few checkpoints die after a few hundred frames but this one looks alive. pic.twitter.com/5Y1wuCVgez
— Mario Klingemann (@quasimondo) January 3, 2018