arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.21402cs.CV
收藏

FlowFixer:面向细节保留的主体驱动生成

FlowFixer: Towards Detail-Preserving Subject-Driven Generation

Amazon(亚马逊公司);Korea University(韩国大学)

Jinyoung Jun, Won-Dong Jang, Wenbin Ouyang, Raghudeep Gadde, Jungbeom Lee

AI总结:

FlowFixer通过图像到图像的直接翻译和自监督训练,提升主体驱动生成中细节保留的保真度和真实性。

摘要:

我们提出了FlowFixer,一种用于主体驱动生成(SDG)的细化框架,旨在恢复生成过程中因主体尺度和视角变化而丢失的细节数字。FlowFixer提出直接从视觉参考进行图像到图像的翻译,避免语言提示中的歧义。为了实现图像到图像的训练,我们引入了一种一步去噪方案来生成自监督训练数据,该方案能够自动去除高频细节同时保留全局结构,从而有效模拟现实中的SDG误差。我们进一步提出基于关键点匹配的度量标准,以在超越CLIP或DINO通常测量的语义相似性之外,正确评估细节的保真度。实验结果表明,FlowFixer在定性和定量评估中均优于最先进的SDG方法,为高保真主体驱动生成设定了新的基准。

英文摘要:

We present FlowFixer, a refinement framework for subject-driven generation (SDG) that restores fine details lost during generation caused by changes in scale and perspective of a subject. FlowFixer proposes direct image-to-image translation from visual references, avoiding ambiguities in language prompts. To enable image-to-image training, we introduce a one-step denoising scheme to generate self-supervised training data, which automatically removes high-frequency details while preserving global structure, effectively simulating real-world SDG errors. We further propose a keypoint matching-based metric to properly assess fidelity in details beyond semantic similarities usually measured by CLIP or DINO. Experimental results demonstrate that FlowFixer outperforms state-of-the-art SDG methods in both qualitative and quantitative evaluations, setting a new benchmark for high-fidelity subject-driven generation.

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。

↑