利用 Diffusion Transformer 免费实现万物个性化
Personalize Anything for Free with Diffusion Transformer
AI总结：
本文提出免训练框架 Personalize Anything，利用 DiT 中的自适应 token 替换与 patch 扰动实现高一致性、多功能的个性化图像生成和编辑。
摘要：
个性化图像生成旨在生成用户指定概念的图像，同时支持灵活编辑。近期的免训练方法虽然比基于训练的方法具有更高的计算效率，但在身份保持、适用性以及与 diffusion transformer（DiT，扩散 Transformer）的兼容性方面存在困难。本文揭示了 DiT 尚未开发的潜力：只需用参考主体的 token 替换去噪 token，即可实现零样本主体重建。这一简单而有效的特征注入技术开启了从个性化到图像编辑的多种场景。基于该观察，我们提出 Personalize Anything，这是一个免训练框架，通过以下方式在 DiT 中实现个性化图像生成：1）时间步自适应 token 替换，通过早期注入保证主体一致性，并通过后期正则化增强灵活性；2）patch 扰动策略以提升结构多样性。我们的方法无缝支持布局引导生成、多主体个性化和遮罩控制编辑。评估表明，其在身份保持和多功能性方面达到最先进性能。我们的工作为 DiT 建立了新的认识，同时为高效个性化提供了实用范式。
英文摘要：
Personalized image generation aims to produce images of user-specified concepts while enabling flexible editing. Recent training-free approaches, while exhibit higher computational efficiency than training-based methods, struggle with identity preservation, applicability, and compatibility with diffusion transformers (DiTs). In this paper, we uncover the untapped potential of DiT, where simply replacing denoising tokens with those of a reference subject achieves zero-shot subject reconstruction. This simple yet effective feature injection technique unlocks diverse scenarios, from personalization to image editing. Building upon this observation, we propose \textbf{Personalize Anything}, a training-free framework that achieves personalized image generation in DiT through: 1) timestep-adaptive token replacement that enforces subject consistency via early-stage injection and enhances flexibility through late-stage regularization, and 2) patch perturbation strategies to boost structural diversity. Our method seamlessly supports layout-guided generation, multi-subject personalization, and mask-controlled editing. Evaluations demonstrate state-of-the-art performance in identity preservation and versatility. Our work establishes new insights into DiTs while delivering a practical paradigm for efficient personalization.
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。
