bryandlee

안녕하세요. 미리 학습된 모델을 사용했다는 것이 어떤 의미인가요? StyleGAN 실험에는 ffhq256 모델을 사용했습니다.

안녕하세요. 결과물 생성을 위해 사용한 코드는 본문의 각 섹션에 링크를 걸어두었습니다. 다만, 데이터나 학습된 모델의 경우 저작권 문제가 있어 공유해드리기는 어려울 것 같습니다. 감사합니다.

The samples can be downloaded using the script: https://github.com/bryandlee/naver-webtoon-data

Hi, the script converts the official stylegan2(tensorflow) checkpoint to a rosinality's stylegan2(pytorch) checkpoint. You can find out more about it [here](https://github.com/rosinality/stylegan2-pytorch#convert-weight-from-official-checkpoints). For the import error, try running the script inside...

Hi, it seems normal for the earlier steps in training.

That's strange. Which dataset are you using? I think the training dataset for stylegan2 generator & encoder should match.

Hi, I didn't keep the intermediate results. Sorry about that :(

Hi, I only qualitatively compared the results with the paper, and it seems like the encoder performance does not match with the original one while the inversion looks ok. ...

This is an unofficial implementation of the paper https://arxiv.org/pdf/2004.00049.pdf. It's an encoder-based latent projection method for GAN generators, which finds a latent code of a given real image.

Hi, the [1, 512]-shaped latent corresponds to the original W space where [1, 14, 512]-shaped latent corresponds to the expanded W+ space. W+ space uses different w vectors for each...