TensorFlow Deep Learning Projects
by Alexey Grigorev, Srinivas Kulkarni, Rajalingappaa Shanmugamani
DCGANs
DCGANs are the first relevant improvement on the GAN architecture. DCGANs always successfully complete their training phase and, given enough epochs and examples, they tend to generate satisfactory quality outputs. That soon made them the baseline for GANs and helped to produce some amazing achievements, such as generating new Pokemon from known ones: https://www.youtube.com/watch?v=rs3aI7bACGc or creating faces of celebrities that actually never existed but are incredibly realistic (nothing uncanny), just as NVIDIA did: https://youtu.be/XOxxPcy5Gr4 using a new training approach called progressing growing: http://research.nvidia.com/sites/default/files/publications/karras2017gan-paper.pdf. They have their root in using the same convolutions ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access