Why in my training, not the same as Latent have the same generator pictures
Opened this issue · 1 comments
zpc333 commented
IIIBlueberry commented
It's called Mode collapse where the GAN generates same images(limited sample variety), despite different latent variables
see here for more info
https://cedar.buffalo.edu/~srihari/CSE676/22.3-GAN%20Mode%20Collapse.pdf
https://developers.google.com/machine-learning/gan/problems#mode-collapse