解决图像生成中物体遮挡顺序模糊问题，让重叠区域更真实
OcclusionFormer: Arranging Z-Order for Layout-Grounded Image Generation
- 用体积渲染显式建模物体前后层次关系
- 在重叠区域减少纹理纠缠，提升遮挡一致性
- 适合需要精确空间布局的图像生成任务
近期布局到图像模型在空间可控性上取得显著进展，但仍面临物体间遮挡关系处理难题。当边界框重叠时，现有方法缺乏显式的遮挡顺序信息，导致交集区域生成歧义，难以确定复杂遮挡关系，常产生纠缠纹理或物理不一致的层叠效果。为此，我们构建了SA-Z数据集，包含大规模显式遮挡排序与像素级标注。基于此，提出OcclusionFormer——一种新型遮挡感知扩散变换框架，通过解耦实例并采用体积渲染进行合成，显式建模Z阶优先级。此外，引入查询对齐损失，显式监督个体实例，增强语义一致性。该方法有效降低重叠区域的歧义性，强制正确的遮挡依赖关系，保持结构完整性，在多种场景下均实现显著精度提升。
原文摘要 · Abstract (English)
Recent layout-to-image models have achieved remarkable progress in spatial controllability. However, they still struggle with inter-object occlusion. When bounding boxes overlap, most existing methods lack explicit occlusion information, which makes the generation in intersection regions inherently ambiguous and hinders the determination of complex occlusion relationships. As a result, they often produce entangled textures or physically inconsistent layering in the overlapped areas. To address this issue, we first construct SA-Z, a large-scale dataset enriched with explicit occlusion ordering and pixel-level annotations. Building upon our proposed dataset, we introduce OcclusionFormer, a novel occlusion-aware Diffusion Transformer framework that explicitly models Z-order priority by decoupling instances and compositing them via volume rendering. Furthermore, to ensure fine-grained spatial precision, we introduce a queried alignment loss that explicitly supervises individual instances and enhances semantic consistency. The proposed method effectively reduces ambiguity in overlapping regions, enforces correct occlusion dependencies, and preserves structural integrity, leading to substantial accuracy gains across diverse scenes.
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。