通过分层渐进关系建模，提升DETR的检测速度与精度。
LP-DETR: Layer-wise Progressive Relations for Object Detection
- 引入可学习的空间关系注意力机制，自适应融合局部、中程和全局关系。
- 在12轮训练内达到52.3% AP，24轮达52.5% AP，使用Swin-L骨干网络时达58.0% AP。
- 早期层聚焦局部关系，深层逐渐转向全局上下文，揭示模型演进规律。
本文提出LP-DETR（分层渐进关系检测器），一种基于DETR的新型目标检测方法，通过多尺度关系建模提升性能。该方法引入可学习的对象查询间空间关系，采用关系感知自注意力机制，自适应地在解码器各层平衡局部、中程和全局关系。这种渐进式设计使模型能有效捕捉检测流程中不断演变的空间依赖性。在COCO 2017数据集上的大量实验表明，相比标准自注意力模块，该方法显著提升收敛速度与检测精度。使用ResNet-50骨干网络时，12轮训练达到52.3% AP，24轮达52.5% AP；采用Swin-L骨干网络时进一步提升至58.0% AP。此外分析发现，模型在早期解码层自然优先关注局部空间关系，随着深度增加逐步转向更广的上下文，为未来目标检测研究提供了重要启示。
原文摘要 · Abstract (English)
This paper presents LP-DETR (Layer-wise Progressive DETR), a novel approach that enhances DETR-based object detection through multi-scale relation modeling. Our method introduces learnable spatial relationships between object queries through a relation-aware self-attention mechanism, which adaptively learns to balance different scales of relations (local, medium and global) across decoder layers. This progressive design enables the model to effectively capture evolving spatial dependencies throughout the detection pipeline. Extensive experiments on COCO 2017 dataset demonstrate that our method improves both convergence speed and detection accuracy compared to standard self-attention module. The proposed method achieves competitive results, reaching 52.3\% AP with 12 epochs and 52.5\% AP with 24 epochs using ResNet-50 backbone, and further improving to 58.0\% AP with Swin-L backbone. Furthermore, our analysis reveals an interesting pattern: the model naturally learns to prioritize local spatial relations in early decoder layers while gradually shifting attention to broader contexts in deeper layers, providing valuable insights for future research in object detection.
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。