search

DmitryUlyanov / AGE

Code for the paper "Adversarial Generator-Encoder Networks"
https://arxiv.org/abs/1704.02304
285 stars 47 forks source link

Inception Score on CelebA #5

Open b03902128 opened 6 years ago

b03902128 commented 6 years ago

Hi,

Just for curiosity, I wonder if I get the right pipeline to check the inception score on CelebA dataset as you presented in the paper.

Should I use the Inception model that is pretrained on ImageNet or trained on CelebA from scratch by myself?

Thanks in advance :)

DmitryUlyanov commented 6 years ago

Hi, we have used a classifier, pretrained on the Imagenet. Note that inception score with such classifier does not make sense. We initially computed such value to compare to other method and removed this evaluation from the most recent version of the paper if I remember correctly.

On Sat, May 5, 2018 at 3:53 PM b03902128 notifications@github.com wrote:

Hi,

Just for curiosity, I wonder if I get the right pipeline to check the inception score on CelebA dataset as you presented in the paper.

Should I use the Inception model that is pretrained on ImageNet or trained on CelebA from scratch by myself?

Thanks in advance :)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/DmitryUlyanov/AGE/issues/5, or mute the thread https://github.com/notifications/unsubscribe-auth/AGanZD0pTY2zwnFOZJYjdWvX9G18iXdTks5tvaDBgaJpZM4TzoEa .

-- Best, Dmitry

b03902128 commented 6 years ago

Hi Dr. Ulyanov ,

Thanks! I found this in the link of the paper in readme, maybe you would like to update the link, too :)

On the other hand, what evaluation protocol would you recommend to show the power of AGE on CelebA dataset? Thank you.