arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2304.14403cs.CVcs.GRcs.LG

Make It So:引导 StyleGAN 实现任意图像反演与编辑

Make It So: Steering StyleGAN for Any Image Inversion and Editing

University of Illinois Urbana-Champaign(伊利诺伊大学厄巴纳-香槟分校)

Anand Bhattad, Viraj Shah, Derek Hoiem, D. A. Forsyth

AI总结:

提出在 $\mathcal{Z}$ 空间运行的 GAN 反演方法 Make It So,可在域外图像上保留编辑能力,并在反演精度和复杂场景编辑质量上显著超越 PTI。

摘要:

StyleGAN 解耦的风格表示通过操纵隐变量实现了强大的图像编辑,但如何将真实世界图像准确映射到其隐变量(即 GAN 反演)仍然是一项挑战。现有的 GAN 反演方法难以同时保持编辑方向并生成逼真结果。为克服这些局限,我们提出 Make It So,这是一种新颖的 GAN 反演方法,运行于 $\mathcal{Z}$(噪声)空间,而非通常使用的 $\mathcal{W}$(隐风格)空间。Make It So 即使对于域外图像也能保留编辑能力,这是以往方法所忽视的关键属性。定量评估表明,Make It So 的反演准确率比最先进方法 PTI 高出五倍,并且在复杂室内场景上实现了十倍的编辑质量提升。

英文摘要:

StyleGAN's disentangled style representation enables powerful image editing by manipulating the latent variables, but accurately mapping real-world images to their latent variables (GAN inversion) remains a challenge. Existing GAN inversion methods struggle to maintain editing directions and produce realistic results. To address these limitations, we propose Make It So, a novel GAN inversion method that operates in the $\mathcal{Z}$ (noise) space rather than the typical $\mathcal{W}$ (latent style) space. Make It So preserves editing capabilities, even for out-of-domain images. This is a crucial property that was overlooked in prior methods. Our quantitative evaluations demonstrate that Make It So outperforms the state-of-the-art method PTI~\cite{roich2021pivotal} by a factor of five in inversion accuracy and achieves ten times better edit quality for complex indoor scenes.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑