Metraipolis › archive › dc_352d4d3dc0ad

GAN-Diff : Coupling Pretrained WGAN-GP Features with Conditional Diffusion U-Nets — arXiv:2608.22272

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.22272
License (as archived)
CC0-1.0
Archived
2026-10-18T07:42:42.162Z
SHA-256 of the archived text
a7b09146a108255643d98ff015b924ceb70e514ecfb75669ae4978d5a58c6cc9
Stable URL of this page
https://metraipolis.com/documento/dc_352d4d3dc0ad

How to cite

GAN-Diff : Coupling Pretrained WGAN-GP Features with Conditional Diffusion U-Nets — arXiv:2608.22272 — arXiv (en), https://arxiv.org/abs/2608.22272 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_352d4d3dc0ad (Metraipolis archive dc_352d4d3dc0ad, sha256 a7b09146a108255643d98ff015b924ceb70e514ecfb75669ae4978d5a58c6cc9)
BibTeX
@misc{dc352d4d3dc0ad,
  title = {GAN-Diff : Coupling Pretrained WGAN-GP Features with Conditional Diffusion U-Nets — arXiv:2608.22272},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.22272},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_352d4d3dc0ad; sha256 a7b09146a108255643d98ff015b924ceb70e514ecfb75669ae4978d5a58c6cc9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.22272. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Generative adversarial networks (GANs) can provide efficient image generation, while diffusion models offer high-quality image restoration but require iterative sampling. This paper presents a hybrid GAN-guided diffusion framework that uses a pretrained Wasserstein GAN with gradient penalty (WGAN-GP) as a feature prior for conditional diffusion-based image restoration. Intermediate features from the frozen WGAN-GP generator are incorporated into a diffusion U-Net through cross-attention and remain fixed during the DDIM sampling process. The framework is evaluated on two restoration tasks, Gaussian denoising and 2Xsuper-resolution, using CelebA face images. During development, several sources of instability were identified and addressed, including adversarial learning-rate imbalance, inappropriate diffusion initialization, excessive corruption, and insufficient parameter averaging. The resulting framework consistently improves the quality of both degraded and low-resolution images. In particular, it improves denoising performance by 4.40 dB in PSNR and super-resolution performance by 3.70 dB over their respective input baselines. These results demonstrate the potential of a frozen GAN feature prior to guide diffusion models toward stable and effective image restoration.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-25 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.22272 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-18T07:37:52.864Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).