it seems that can't get noise with your code,
it return none. how do you get it ?

thanks

Help me! Since the paper: "Image2StyleGAN: How to Embed Images Into the StyleGAN Latent Space?" don't have an official implement and in the paper 3.3 just said:

W+ is a concatenation of 18 different 512-dimensional w vectors

so? W space is already 1x18x512，what's the difference between them?

In your inverse.py, I see the w is get by a mapping pretrained net , which convert U(0, 1) to W space or W+ maybe?

Hi, I got an error for this part

np.save("/content/drive/MyDrive/GAN-inversion/latent_w/noise/{}.npy".format(name),np.array(noises))