Any specifications on uploading an image for interpolation? - Githubissues

genforce / idinvert_pytorch

[ECCV 2020] In-Domain GAN Inversion for Real Image Editing (PyTorch code)

https://genforce.github.io/idinvert/

MIT License

407 stars 54 forks source link

Any specifications on uploading an image for interpolation? #10

Closed JettChenT closed 3 years ago

JettChenT commented 3 years ago

While I was running your notebook for interpolation, I encountered this issue: I uploaded this image as the target image and selected the default image for source image: After performing interpolation, this happened: 9k (1) No errors were reported throughout the process, and a similar result was produced upon trying different png files. Any idea on what went wrong?

zhujiapeng commented 3 years ago

This is the failure case caused by the face detector.

gsygsy96 commented 3 years ago

@JettChenT , could you please show the aligned target images?

JettChenT commented 3 years ago

@shyanguan thank you for responding. I'm new to this so could you please tell me how to do that?

gsygsy96 commented 3 years ago

The input image should be aligned first. Have you done this step?

JettChenT commented 3 years ago

@shyanguan No. How do I align the input image?

gsygsy96 commented 3 years ago

Refer to another repo, named StyleGAN-encoder.

JettChenT commented 3 years ago

@ShenYujun Should the aligned image look like this? IMG_1988_01

ShenYujun commented 3 years ago

OMG, it should not look like this. Is your image with three channels? I think it happens because the pixels misalign across channels.

JettChenT commented 3 years ago

Thanks a lot for the support! It really helped.