Skip to main content
QUICK REVIEW

[论文解读] UnionDet: Union-Level Detector Towards Real-Time Human-Object Interaction Detection

Bumsoo Kim, Tae-Ho Choi|arXiv (Cornell University)|Dec 19, 2023
Advanced Neural Network Applications被引用 4
一句话总结

本文提出UnionDet,一种单阶段人体-物体交互(HOI)检测框架,通过一种新颖的联合区域检测器直接预测交互区域(人体-物体对的并集),消除了对后续目标检测配对的需求。通过引入联合区域锚点标注、目标物体分类损失和前景焦点损失,UnionDet在V-COCO和HICO-DET数据集上超越了最先进方法,同时将交互预测的推理速度提升了4×–14×。

ABSTRACT

Recent advances in deep neural networks have achieved significant progress in detecting individual objects from an image. However, object detection is not sufficient to fully understand a visual scene. Towards a deeper visual understanding, the interactions between objects, especially humans and objects are essential. Most prior works have obtained this information with a bottom-up approach, where the objects are first detected and the interactions are predicted sequentially by pairing the objects. This is a major bottleneck in HOI detection inference time. To tackle this problem, we propose UnionDet, a one-stage meta-architecture for HOI detection powered by a novel union-level detector that eliminates this additional inference stage by directly capturing the region of interaction. Our one-stage detector for human-object interaction shows a significant reduction in interaction prediction time 4x~14x while outperforming state-of-the-art methods on two public datasets: V-COCO and HICO-DET.

研究动机与目标

  • 解决多阶段HOI检测器效率低下的问题,其需要按顺序进行目标检测和交互配对,造成主要的推理瓶颈。
  • 通过消除对物体对上繁重的后处理推理,实现真正的实时HOI检测。
  • 开发一种统一的、端到端可训练的单阶段架构,同时检测物体和交互关系。
  • 克服联合区域检测面临的技术挑战,如对人类区域的偏差以及基于IoU的匹配不准确问题。
  • 在标准目标检测的额外推理时间极低的前提下,实现最先进性能。

提出的方法

  • 提出一种新颖的联合区域检测器,直接回归交互区域(人体与物体的并集),而非依赖检测后对物体的配对。
  • 引入联合区域锚点标注,基于联合区域级别的IoU匹配将锚点分配给交互区域,提升定位精度。
  • 设计目标物体分类损失($\mathcal{L}^{cls}_{ij}$),以更好地区分交互对中的物体类别。
  • 开发联合前景焦点损失(FFL),根据联合区域级别的置信度重新加权正样本,减轻背景偏差。
  • 使用联合匹配函数($\mu_u$),用更鲁棒的度量替代标准IoU,以匹配锚点与交互区域。
  • 将联合检测器集成到元架构(UnionDet)中,兼容现有单阶段目标检测器,支持端到端训练。
Figure 1: Technical Challenges in Union Detection. (Left) The box with the highest confidence for the ground-truth action ’throw’ is highlighted in bold, and the region with the highest IoU with the ground-truth union region is dotted. As you can see, the highest confidence is biased towards the hum
Figure 1: Technical Challenges in Union Detection. (Left) The box with the highest confidence for the ground-truth action ’throw’ is highlighted in bold, and the region with the highest IoU with the ground-truth union region is dotted. As you can see, the highest confidence is biased towards the hum

实验结果

研究问题

  • RQ1单阶段检测器能否在不依赖独立配对阶段的前提下,直接预测人体-物体交互区域?
  • RQ2联合区域检测如何克服对人类区域的偏差以及基于IoU匹配不准确的挑战?
  • RQ3需要哪些训练技术来稳定并提升联合区域检测的性能?
  • RQ4与多阶段HOI检测器相比,联合区域检测器在推理时间上能减少多少?
  • RQ5联合区域检测器能否在保持实时推理速度的同时实现最先进性能?

主要发现

  • 与基线多阶段HOI检测器相比,UnionDet在交互预测上的推理速度提升了4×–14×,后者在目标检测后每张图像需40ms–130ms。
  • 该模型在标准目标检测器基础上,将额外推理时间减少至仅9.06ms(占11.7%),显著优于先前方法。
  • 在V-COCO数据集上,UnionDet在两种评估场景下分别达到47.5%和56.2%的平均精度,超越最先进方法。
  • 在HICO-DET数据集上,UnionDet在“默认”和“已知物体”设置下均达到最先进性能,且无需依赖外部特征。
  • 消融实验证实,每个组件——联合分支、目标物体分类损失、联合匹配函数和前景焦点损失——均显著提升了检测性能。
  • 该模型在保持完全端到端可训练性的同时,兼容现有单阶段目标检测器,支持实时部署。
Figure 2: The overall architecture of UnionDet. Our UnionDet is generally compatible with one-stage object detectors. The feature pyramid obtained from the backbone network is simultaneously fed to Union Branch and Instance Branch. While Union Branch directly captures the region of interaction, Inst
Figure 2: The overall architecture of UnionDet. Our UnionDet is generally compatible with one-stage object detectors. The feature pyramid obtained from the backbone network is simultaneously fed to Union Branch and Instance Branch. While Union Branch directly captures the region of interaction, Inst

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。