search

genforce / interfacegan

[CVPR 2020] Interpreting the Latent Space of GANs for Semantic Face Editing
https://genforce.github.io/interfacegan/
MIT License
1.51k stars 281 forks source link

Latent code- attribute score pair #53

Closed kLamda closed 3 years ago

kLamda commented 4 years ago

What are the latent code-attribute score as mentioned in the train_boundary.py file in the repository? How can we calculate the attribute scores?

miss1997yuan commented 4 years ago

add one

ShenYujun commented 4 years ago

Use a pre-trained attribute classifier. We may provide some classifiers in the future. You can also train your own classifiers.

kLamda commented 4 years ago

Yeah, Can you provide those classifiers shortly or direct us the way to generate customized classifiers ? It would be a great help.

RioRic commented 3 years ago

When you can share the per-tained attribute classier or code to generate the classifier

yoichi1484 commented 3 years ago

Can you mind telling me what dataset you used for the attribute classifier?

ShenYujun commented 3 years ago

It is the CelebA data that we use. You can find the link here.

yoichi1484 commented 3 years ago

It is the CelebA data that we use. You can find the link here.

Thank you!

liujingwen-bmil commented 8 months ago

Yeah, Can you provide those classifiers shortly or direct us the way to generate customized classifiers ? It would be a great help.

hello, have you already known how to generated customized classifiers? Any help would be appreciated!

YKJamesMoriarty commented 2 months ago

潜在空间就是stylegan生成器的输入，这个输入是从高维正态分布中取值的一个向量X。然后用这个向量X作为输入在已经训练好的stylegan中推理生成一张图片，接着将这张图片送入一个与训练好的某些人像特征的分类器（如年龄、微笑程度、是否带了眼镜），得到这张图片中这几个特征的分数，最后将生成这种图片的向量X与得到的分数作为一对信息，传入SVM中做语义边界划分