arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2408.01664cs.CVcs.AI
收藏

SAT3D:3D 中的图像驱动语义属性迁移

SAT3D: Image-driven Semantic Attribute Transfer in 3D

Wuhan University(武汉大学);The University of Hong Kong(香港大学);Huawei Technologies(华为技术有限公司)

Zhijun Zhai, Zengmao Wang, Xiaoxiao Long, Kaixuan Zhou, Bo Du

AI总结:

本文提出 SAT3D,在 3D 感知 StyleGAN 风格空间中学习语义属性与风格通道关联,并利用 CLIP 驱动的 QMM 损失实现参考图像属性的摄影级 3D 迁移。

摘要:

基于 GAN 的图像编辑任务旨在在生成模型的潜空间中操控图像属性。以往大多数 2D 和具有 3D 感知能力的方法主要关注编辑图像中语义模糊的属性,或来自参考图像的区域,因而无法实现摄影级语义属性迁移,例如从一张男性照片中迁移胡须。本文提出一种图像驱动的 3D 语义属性迁移方法(SAT3D),通过编辑来自参考图像的语义属性来实现。对于所提方法,探索在基于 StyleGAN 的预训练 3D 感知生成器的风格空间中进行,方式是学习语义属性与风格编码通道之间的相关性。为提供引导,我们将每个属性与一组基于短语的描述符组相关联,并开发了定量测量模块(QMM),以基于描述符组定量描述图像中的属性特征,该模块利用了 CLIP 的图文理解能力。在训练过程中,QMM 被纳入属性损失,以计算图像之间的属性相似性,从而引导目标语义迁移和无关语义保持。我们展示了跨多个领域的 3D 感知属性迁移结果,并与经典 2D 图像编辑方法进行了比较,证明了 SAT3D 的有效性和可定制性。

英文摘要:

GAN-based image editing task aims at manipulating image attributes in the latent space of generative models. Most of the previous 2D and 3D-aware approaches mainly focus on editing attributes in images with ambiguous semantics or regions from a reference image, which fail to achieve photographic semantic attribute transfer, such as the beard from a photo of a man. In this paper, we propose an image-driven Semantic Attribute Transfer method in 3D (SAT3D) by editing semantic attributes from a reference image. For the proposed method, the exploration is conducted in the style space of a pre-trained 3D-aware StyleGAN-based generator by learning the correlations between semantic attributes and style code channels. For guidance, we associate each attribute with a set of phrase-based descriptor groups, and develop a Quantitative Measurement Module (QMM) to quantitatively describe the attribute characteristics in images based on descriptor groups, which leverages the image-text comprehension capability of CLIP. During the training process, the QMM is incorporated into attribute losses to calculate attribute similarity between images, guiding target semantic transferring and irrelevant semantics preserving. We present our 3D-aware attribute transfer results across multiple domains and also conduct comparisons with classical 2D image editing methods, demonstrating the effectiveness and customizability of our SAT3D.

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。

↑