轻量级模块提升视觉模型多尺度特征融合，不增加推理开销。
P$^2$HCT: Plug-and-Play Hierarchical C2F Transformer for Multi-Scale Feature Fusion
- 分层粗到细令牌选择+共享注意力参数，降低计算开销。
- 在YOLOv11上提升mAP 0.9%~0.4%，ImageNet准确率提升6.5%~1.0%。
- 可即插即用，无需重训练，适合边缘设备部署。
特征融合在视觉模型中至关重要，但现有基于注意力的融合方法常伴随巨大计算开销和实现复杂性，尤其在资源受限场景下。为此，我们提出轻量级的即插即用分层C2F Transformer（P$^2$HCT），结合粗到细令牌选择与共享注意力参数，在保留空间细节的同时降低推理成本。P$^2$HCT仅需粗粒度注意力即可训练，推理时可无缝激活以提升精度而无需重训练。将其集成至YOLOv11-N/S/M等实时检测器，在MS COCO上分别获得0.9%、0.5%、0.4%的mAP提升，延迟增加极小。嵌入ResNet-18/50/101骨干网络后，ImageNet top-1准确率分别提升6.5%、1.7%、1.0%。结果表明，P$^2$HCT是硬件友好且通用的检测与分类增强方案。
原文摘要 · Abstract (English)
Feature fusion plays a pivotal role in achieving high performance in vision models, yet existing attention-based fusion techniques often suffer from substantial computational overhead and implementation complexity, particularly in resource-constrained settings. To address these limitations, we introduce the Plug-and-Play Hierarchical C2F Transformer (P$^2$HCT), a lightweight module that combines coarse-to-fine token selection with shared attention parameters to preserve spatial details while reducing inference cost. P$^2$HCT is trainable using coarse attention alone and can be seamlessly activated at inference to enhance accuracy without retraining. Integrated into real-time detectors such as YOLOv11-N/S/M, P$^2$HCT achieves mAP gains of 0.9\%, 0.5\%, and 0.4\% on MS COCO with minimal latency increase. Similarly, embedding P$^2$HCT into ResNet-18/50/101 backbones improves ImageNet top-1 accuracy by 6.5\%, 1.7\%, and 1.0\%, respectively. These results underscore P$^2$HCT's effectiveness as a hardware-friendly and general-purpose enhancement for both detection and classification tasks.
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。