Awesome

BEGAN-torch7

Torch7 implementation of BEGAN: Boundary Equilibrium Generative Adversarial Networks

Note that this project is in progress

Results so far

• training is unstable yet. It seems to converge at the very early stage, but finally diverge in the end.
• The max value of kt set as 0.01 due to avoiding modal collapse (the paper used 1.0)

Author

Minchul Shin, @nashory