Controllable Multi-Object Insertion for Detection Data Augmentation

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.26922 (cs) [Submitted on 17 Oct 2026] Title: Controllable Multi-Object Insertion for Detection Data Augmentation Authors: Jiyang Huang, Tianwei Zhang, Xiaoqin Zhang, Ling Shao, Shijian Lu.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.26922 (cs) [提交于 2026 年 10 月 17 日] 标题：用于检测数据增强的可控多目标插入 作者：Jiyang Huang, Tianwei Zhang, Xiaoqin Zhang, Ling Shao, Shijian Lu。

Abstract: Controllable object insertion provides a targeted approach to augmenting object detection datasets that require category-count fidelity, annotation alignment, and source-content preservation beyond visual realism. Several approaches have been proposed to address this grand challenge, but most suffer from different constraints in precise category-count requests and visual consistency, limited coordination across categories and instances, and poor localization of multiple objects of specific category and count.

摘要：可控目标插入为增强目标检测数据集提供了一种有针对性的方法，这些数据集在视觉真实感之外，还需要类别计数保真度、标注对齐以及源内容保留。目前已提出多种方法来应对这一重大挑战，但大多数方法在精确的类别计数请求和视觉一致性方面存在不同限制，在跨类别和实例的协调方面能力有限，且难以对特定类别和数量的多个目标进行准确的定位。

We design SceneP2F, a two-stage trainable framework that connects scene-aware joint planning with layout-aligned generation leveraging explicit instance-level layouts. SceneP2F consists of two novel designs. The first is ScenePlan which allocates one class-labeled query per requested instance, jointly predicts scene-compatible object centers, and separately estimates box sizes conditioned on the planned centers and scene context.

我们设计了 SceneP2F，这是一个两阶段可训练框架，通过利用显式的实例级布局，将场景感知联合规划与布局对齐生成连接起来。SceneP2F 包含两个创新设计。第一个是 ScenePlan，它为每个请求的实例分配一个带有类别标签的查询，联合预测与场景兼容的目标中心，并根据规划的中心和场景上下文分别估计框的大小。

The second is SceneAlign-Fill which combines geometry- and appearance-aware conditioning, box-class-guided regional attention, and an annotation-region-weighted flow-matching objective to render the planned objects while limiting unintended changes to source content. Extensive experiments on three challenging benchmarks UAVDT, VisDrone, and BDD100K show that SceneP2F outperforms representative general-purpose editors and object-insertion methods consistently, demonstrating the value of coordinating instance-level planning and generation for category-targeted detection augmentation.

第二个是 SceneAlign-Fill，它结合了几何和外观感知调节、框类引导的区域注意力机制，以及标注区域加权的流匹配目标，在渲染规划目标的同时限制对源内容的意外更改。在 UAVDT、VisDrone 和 BDD100K 三个具有挑战性的基准测试上的大量实验表明，SceneP2F 始终优于代表性的通用编辑器和目标插入方法，证明了在类别定向检测增强中协调实例级规划与生成的价值。