面向无人机目标检测的视觉原型条件化焦点区域生成
Visual Prototype Conditioned Focal Region Generation for UAV-Based Object Detection
用类别视觉原型作为扩散模型的条件嵌入，替代纯布局框条件，可缓解小目标边界伪影；配套的焦点区域增强与标签精炼流程可直接用于任何检测器的合成数据增广。
原文依据 UAVGen designs a Visual Prototype Conditioned Diffusion Model (VPC-DM) that constructs representative instances for each class
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论基于无人机的目标检测在标注训练数据有限的动态变化场景中是一项关键但具有挑战性的任务。布局到图像生成方法已被证明能够通过基于扩散模型合成带标注图像来提升检测精度。然而，这类方法常常产生伪影，尤其是在微小目标的布局边界附近，从而严重限制了其性能。为解决这些问题，我们提出 UAVGen，一种面向无人机目标检测定制的布局到图像生成框架。具体而言，UAVGen 设计了视觉原型条件扩散模型（VPC-DM），为每个类别构建代表性实例并将其整合进潜变量嵌入中，以实现高保真的目标生成。此外，我们引入焦点区域增强数据流水线（FRE-DP），在合成中强调目标集中的前景区域，并结合标签精炼来纠正缺失、多余与错位的生成。大量实验结果表明，我们的方法显著优于最优方法，并在与不同检测器集成时持续提升精度。源代码见 https://github.com/Sirius-Li/UAVGen。
英文原文摘要 1313 字符
Unmanned aerial vehicle (UAV) based object detection is a critical but challenging task, when applied in dynamically changing scenarios with limited annotated training data. Layout-to-image generation approaches have proved effective in promoting detection accuracy by synthesizing labeled images based on diffusion models. However, they suffer from frequently producing artifacts, especially near layout boundaries of tiny objects, thus substantially limiting their performance. To address these issues, we propose UAVGen, a novel layout-to-image generation framework tailored for UAV-based object detection. Specifically, UAVGen designs a Visual Prototype Conditioned Diffusion Model (VPC-DM) that constructs representative instances for each class and integrates them into latent embeddings for high-fidelity object generation. Moreover, a Focal Region Enhanced Data Pipeline (FRE-DP) is introduced to emphasize object-concentrated foreground regions in synthesis, combined with a label refinement to correct missing, extra and misaligned generations. Extensive experimental results demonstrate that our method significantly outperforms state-of-the-art approaches, and consistently promotes accuracy when integrated with distinct detectors. The source code is available at https://github.com/Sirius-Li/UAVGen.