a312863063/seeprettyface-face_editor

这是一个基于StyleGAN的人脸属性编辑器

seeprettyface-face_editor's Issues

play_with_dlatent.py StopIteration

大神 求帮助

Traceback (most recent call last):
File "play_with_dlatent.py", line 87, in
main()
File "play_with_dlatent.py", line 44, in main
generator = Generator(Gs_network, batch_size=1, randomize_noise=False)
File "E:\stylegan2\stylegan2_application\humanfacegenerators\util\generator_model.py", line 52, in init
self.dlatent_variable = next(v for v in tf.global_variables() if 'learnable_dlatents' in v.name)
StopIteration

能否完成大幅度的角度改变

您好，我想问下这个水平或竖直方向改变能否完成大幅度的角度改变（比如把正脸调成侧脸）。因为我看您放的示例图只有轻微的角度改变。

有关colab上测试报错G_synthesis/_Run/concat:0的建议

在colab上测试。

报错是：
……
KeyError: "The name 'G_synthesis_1/_Run/concat:0' refers to a Tensor which does not exist. The operation, 'G_synthesis_1/_Run/concat', does not exist in the graph."

During handling of the above exception, another exception occurred:

……
KeyError: "The name 'G_synthesis/_Run/concat:0' refers to a Tensor which does not exist. The operation, 'G_synthesis/_Run/concat', does not exist in the graph."

请遇到这个问题的小伙伴统一修改unit目录中的generator_model.py文件
将generator类中：
try:
self.generator_output = self.graph.get_tensor_by_name('G_synthesis_1/_Run/concat:0')
except KeyError:
# If we loaded only Gs and didn't load G or D, then scope "G_synthesis_1" won't exist in the graph.
self.generator_output = self.graph.get_tensor_by_name('G_synthesis/_Run/concat:0')

修改为：
try:
self.generator_output = self.graph.get_tensor_by_name('G_synthesis_1/_Run/concat/concat:0')
except KeyError:
# If we loaded only Gs and didn't load G or D, then scope "G_synthesis_1" won't exist in the graph.
self.generator_output = self.graph.get_tensor_by_name('G_synthesis/_Run/concat:0')

供楼主参考。

python play_with_dlatent.py 报错～

harry@MacBook-Pro seeprettyface % python play_with_dlatent.py             
  File "play_with_dlatent.py", line 9
SyntaxError: Non-ASCII character '\xe4' in file play_with_dlatent.py on line 9, but no encoding declared; see http://python.org/dev/peps/pep-0263/ for details

对于任意输入图片获取latent的方法

我们现在使用的是 pbaylies/stylegan-encoder 的方法，但是获取一张图片的latent太慢了。看您的博客曾经介绍过BiGAN的方法获取latent，不知效果如何。

能有头发的吗？

眼睛的视角能有变化吗

使用生成器的生成图片的数据集都是视线正对着的画面的，想请问有什么方法可以做到调整眼睛的位置

play_with_dlatent.py AssertionError

Traceback (most recent call last):
File "play_with_dlatent.py", line 87, in
main()
File "play_with_dlatent.py", line 43, in main
generator_network, discriminator_network, Gs_network = pickle.load(f)
File "D:\MyProjects\ai\seeprettyface-face_editor\dnnlib\tflib\network.py", line 279, in setstate
assert state["version"] in [2, 3]
AssertionError

what's happening ?

TypeError: zip argument #1 must support iteration

stylegan生成中加入的noise是如何控制的

latent_directions

您好！请问 latent_directions 中 npy 文件的训练代码有开源吗？非常感谢！

请问潜码的txt文件时如何生成的呢？

how to get emotion/gender.npy??

Excellent work!
an issue is how you get emotion.npy or other *npy?

Where is the link for downloading 'model/generator_yellow.pkl'?

Hi,
I can only find the model in stylegan2 but not stylegan. Where should I look for?
Thanks!

是否可以用自己的照片去进行属性修改？

作者您好，我看了一下代码，里面是通过已经训练好的模型然后随机生成出来的图片(png)和隐向量(.txt)文件，然后可以进行属性修改，表现效果还是很好的，那如果我就用我的一张自拍照，属性修改怎么处理呢？期待您的解答，感谢！

权重文件

权重文件在那下载？

An error about ''generator_yellow.pkl''

KeyError: "The name 'G_synthesis/_Run/concat:0' refers to a Tensor which does not exist. The operation, 'G_synthesis/_Run/concat', does not exist in the graph."

Generating latent spaces for our own images.

Hi @a312863063 ,

Thank you for this great repository and we are using this for our GANs research.

I need a clarification for creating the latent space vector from our own image to feed into this model. I can able to create new image and using that image can able to adjust the different attributes using latest directions like smile, age, etc.

I tried StyleGAN-encoder package to create latent space with FFHQ model and its size is (18,512) with *.npy format. But this repository needs in the format of .txt and its length is around (512). If possible can you please help me on how can I generate the latent space for the original(our own) images which is compatible with this model(i.e. latent spaces in *.txt format without any loss in the quality of image).

Also, while we create latent space form real image using StyleGan encoder, we are facing some loss in the quality, is it because of the model or do I need to adjust any parameter in general?

Thanks in advance, looking forward you response.