V-Shuffle: Zero-Shot Style Transfer via Value Shuffle

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tang, Haojun, Lin, Qiwei, Xu, Tongda, Huang, Lida, Wang, Yan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917069740572672
author Tang, Haojun
Lin, Qiwei
Xu, Tongda
Huang, Lida
Wang, Yan
author_facet Tang, Haojun
Lin, Qiwei
Xu, Tongda
Huang, Lida
Wang, Yan
contents Attention injection-based style transfer has achieved remarkable progress in recent years. However, existing methods often suffer from content leakage, where the undesired semantic content of the style image mistakenly appears in the stylized output. In this paper, we propose V-Shuffle, a zero-shot style transfer method that leverages multiple style images from the same style domain to effectively navigate the trade-off between content preservation and style fidelity. V-Shuffle implicitly disrupts the semantic content of the style images by shuffling the value features within the self-attention layers of the diffusion model, thereby preserving low-level style representations. We further introduce a Hybrid Style Regularization that complements these low-level representations with high-level style textures to enhance style fidelity. Empirical results demonstrate that V-Shuffle achieves excellent performance when utilizing multiple style images. Moreover, when applied to a single style image, V-Shuffle outperforms previous state-of-the-art methods.
format Preprint
id arxiv_https___arxiv_org_abs_2511_06365
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle V-Shuffle: Zero-Shot Style Transfer via Value Shuffle
Tang, Haojun
Lin, Qiwei
Xu, Tongda
Huang, Lida
Wang, Yan
Computer Vision and Pattern Recognition
Attention injection-based style transfer has achieved remarkable progress in recent years. However, existing methods often suffer from content leakage, where the undesired semantic content of the style image mistakenly appears in the stylized output. In this paper, we propose V-Shuffle, a zero-shot style transfer method that leverages multiple style images from the same style domain to effectively navigate the trade-off between content preservation and style fidelity. V-Shuffle implicitly disrupts the semantic content of the style images by shuffling the value features within the self-attention layers of the diffusion model, thereby preserving low-level style representations. We further introduce a Hybrid Style Regularization that complements these low-level representations with high-level style textures to enhance style fidelity. Empirical results demonstrate that V-Shuffle achieves excellent performance when utilizing multiple style images. Moreover, when applied to a single style image, V-Shuffle outperforms previous state-of-the-art methods.
title V-Shuffle: Zero-Shot Style Transfer via Value Shuffle
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.06365