search

jerryli27 / TwinGAN

Twin-GAN -- Unpaired Cross-Domain Image Translation with Weight-Sharing GANs
Apache License 2.0
719 stars 98 forks source link

identity preserving? #17

Open Yijunmaverick opened 5 years ago

Yijunmaverick commented 5 years ago

Thanks for your great work. I'm just wondering how do you measure or evaluate how well the identity is preserved in your results?

At least visually it is hard for me to recognize Emma in the anime results.

jerryli27 commented 5 years ago

You're correct. I only measured identity preservation qualitatively by basically eyeballing and it's only a proof of concept with a lot of room for improvements.

An indirect way is to measure the distance between the embedding of the original image versus the generated anime version -- which we can track from training logs. Maybe I should include that number somewhere in the paper. Thanks for the suggestion!