无需训练即可实现精准图像几何编辑，支持复杂变形。
Training-free Geometric Image Editing on Diffusion Models
- 分步解耦编辑流程：变换、修复、精修三阶段独立处理
- 在GeoBench上相比现有方法，保真度与编辑精度显著提升
- 适用于大尺度或复杂结构的图像重排，适合视觉创作场景
我们研究几何图像编辑任务，即在保持场景整体一致性的同时，对图像中物体进行重定位、重定向或重塑。以往基于扩散模型的方法通常试图一步完成所有子任务，在变换较大或结构复杂时效果不佳。为此，我们提出一种解耦式流水线，将物体变换、源区域修复和目标区域精修分开处理。修复与精修均采用无需训练的扩散方法FreeFine。在新提出的GeoBench基准测试中，该方法涵盖2D与3D编辑场景，相比前沿方法在图像保真度与编辑精度上表现更优，尤其在高挑战性变换下优势明显。代码与数据集已开源。
原文摘要 · Abstract (English)
We tackle the task of geometric image editing, where an object within an image is repositioned, reoriented, or reshaped while preserving overall scene coherence. Previous diffusion-based editing methods often attempt to handle all relevant subtasks in a single step, proving difficult when transformations become large or structurally complex. We address this by proposing a decoupled pipeline that separates object transformation, source region inpainting, and target region refinement. Both inpainting and refinement are implemented using a training-free diffusion approach, FreeFine. In experiments on our new GeoBench benchmark, which contains both 2D and 3D editing scenarios, FreeFine outperforms state-of-the-art alternatives in image fidelity, and edit precision, especially under demanding transformations. Code and benchmark are available at: https://github.com/CIawevy/FreeFine
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。