Papers › Delving StyleGAN Inversion for Image Editing: A Foundation Latent Space Viewpoint

Delving StyleGAN Inversion for Image Editing: A Foundation Latent Space Viewpoint

21 Nov 2022CVPR 2023 1arXiv:2211.11448archive 2025-07-28

Hongyu Liu, Yibing Song, Qifeng Chen

GAN inversion and editing via StyleGAN maps an input image into the embedding spaces (𝒲, 𝒲^+, and ℱ) to simultaneously maintain image fidelity and meaningful manipulation. From latent space 𝒲 to extended latent space 𝒲^+ to feature space ℱ in StyleGAN, the editability of GAN inversion decreases while its reconstruction quality increases. Recent GAN inversion methods typically explore 𝒲^+ and ℱ rather than 𝒲 to improve reconstruction fidelity while maintaining editability. As 𝒲^+ and ℱ are derived from 𝒲 that is essentially the foundation latent space of StyleGAN, these GAN inversion methods focusing on 𝒲^+ and ℱ spaces could be improved by stepping back to 𝒲. In this work, we propose to first obtain the precise latent code in foundation latent space 𝒲. We introduce contrastive learning to align 𝒲 and the image space for precise latent code discovery. %The obtaining process is by using contrastive learning to align 𝒲 and the image space. Then, we leverage a cross-attention encoder to transform the obtained latent code in 𝒲 into 𝒲^+ and ℱ, accordingly. Our experiments show that our exploration of the foundation latent space 𝒲 improves the representation ability of latent codes in 𝒲^+ and features in ℱ, which yields state-of-the-art reconstruction fidelity and editability results on the standard benchmarks. Project page: https://kumapowerliu.github.io/CLCAE.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

kumapowerliu/clcae officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Contrastive Learning

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

ALIGNAdaptive Instance NormalizationContrastive LearningConvolutionDense ConnectionsFeedforward NetworkR1 RegularizationStyleGAN

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections