search

genforce / sefa

[CVPR 2021] Closed-Form Factorization of Latent Semantics in GANs
https://genforce.github.io/sefa/
MIT License
964 stars 108 forks source link

About attribute predictor and re-scoring analysis. #12

Open GreenLimeSia opened 3 years ago

GreenLimeSia commented 3 years ago

Dear @ShenYujun :

Thanks for your excellent works. It is very useful for my research and I will cite your work in my study. However, I have some issues with re-scoring analysis since the attribute predictor is lost (The same issue on InterfaceGAN). Would you mind releasing your pre-trained attribute predictor's code and model?

Thank you again. Best wishes.

ShenYujun commented 3 years ago

Please refer to this repo.

GreenLimeSia commented 3 years ago

Many thanks for your prompt reply. After reading this repo, I still haven't found how to train a predictor with an attribute of the pose. As you noted in the paper, do you use the 5-point facial landmarks? If so, can you give me some advice on how to take advantage of these key points as pose?

ShenYujun commented 3 years ago

A very simple solution as an example: to get the pose (i.e., left v.s. right), you can use the x-coordinate of the nose point or the x-coordinate of the middle point between two eyes.

GreenLimeSia commented 3 years ago

Sure, I am very grateful for your guidance. I will train a classifier for it.

di-mi-ta commented 3 years ago

Hi @ShenYujun,

Could you release code for rescoring-analysis metric in order to reproduce the results in the paper?

Thanks!