Improving generative adversarial network inversion via fine-tuning GAN encoders
Fuente:
arXiv
Salvato in:
| Autori principali: | , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2021
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866913608330379264 |
|---|---|
| author | Yu, Cheng Wang, Wenmin Bugiolacchi, Roberto |
| author_facet | Yu, Cheng Wang, Wenmin Bugiolacchi, Roberto |
| contents | Generative adversarial networks (GANs) can synthesize high-quality (HQ) images, and GAN inversion is a technique that discovers how to invert given images back to latent space. While existing methods perform on StyleGAN inversion, they have limited performance and are not generalized to different GANs. To address these issues, we proposed a self-supervised method to pre-train and fine-tune GAN encoders. First, we designed an adaptive block to fit different encoder architectures for inverting diverse GANs. Then we pre-train GAN encoders using synthesized images and emphasize local regions through cropping images. Finally, we fine-tune the pre-trained GAN encoder for inverting real images. Compared with state-of-the-art methods, our method achieved better results that reconstructed high-quality images on mainstream GANs. Our code and pre-trained models are available at: https://github.com/disanda/Deep-GAN-Encoders. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2108_10201 |
| institution | arXiv |
| publishDate | 2021 |
| record_format | arxiv |
| spellingShingle | Improving generative adversarial network inversion via fine-tuning GAN encoders Yu, Cheng Wang, Wenmin Bugiolacchi, Roberto Computer Vision and Pattern Recognition Image and Video Processing Generative adversarial networks (GANs) can synthesize high-quality (HQ) images, and GAN inversion is a technique that discovers how to invert given images back to latent space. While existing methods perform on StyleGAN inversion, they have limited performance and are not generalized to different GANs. To address these issues, we proposed a self-supervised method to pre-train and fine-tune GAN encoders. First, we designed an adaptive block to fit different encoder architectures for inverting diverse GANs. Then we pre-train GAN encoders using synthesized images and emphasize local regions through cropping images. Finally, we fine-tune the pre-trained GAN encoder for inverting real images. Compared with state-of-the-art methods, our method achieved better results that reconstructed high-quality images on mainstream GANs. Our code and pre-trained models are available at: https://github.com/disanda/Deep-GAN-Encoders. |
| title | Improving generative adversarial network inversion via fine-tuning GAN encoders |
| topic | Computer Vision and Pattern Recognition Image and Video Processing |
| url | https://arxiv.org/abs/2108.10201 |