[论文] Occlusion for Free: Detecting Occlusion from the Disagreement Between ...

论文概要

研究领域: CV
作者: Chibueze Praise Onyenweaku, Toyibat Adele, Sikirulahi Opeyemi Abdulkareem, Obed Honour Eje
发布时间: 2026-11-04
arXiv: 2611.06899

中文摘要

即使检测器无法恢复被遮挡像素，知道行人被遮挡对机器人仍有价值。但遮挡估计器通常在部署环境很少具备的遮挡标注上训练。我们表明，遮挡可从仅摄像头管线已计算的两个距离估计中读出：表观高度在遮挡下基本不变（检测框仍覆盖整个人）；单目度量深度被拉向遮挡物（深度图报告框内最近表面）。只有排序才重要——只要深度模型误差接近纯增益（无需标签即可检验此条件），就不需要焦距、身高或深度尺度。该分数在 KITTI 上检测重度遮挡 AUC 0.909（全部标注行人 0.888，含检测器遗漏者），无需训练、无需遮挡标签，且非距离代理变量：固定距离区间内 AUC 0.894-0.925，而仅真实距离接近随机。CARLA 合成场景（实测可见度）达 0.819，仅距离为 0.644。我们是在测量三种标准架构对遮挡的响应（深度融合、遮挡模拟增强、注意力——测试配置均未在训练分布外改进检测），并发现德国道路数据训练的检测器在尼日利亚拉各斯自标注街景中精度损失一半之后得出此方法的。基准未代表的环境中，无需本地标注的信号价值更高。

原文摘要

Knowing that a pedestrian is occluded is useful to a robot even when the detector cannot recover the hidden pixels, but occlusion estimators are normally trained on occlusion annotations that deployment settings rarely have. We show that occlusion can be read off two distance estimates a camera-only pipeline already computes. Apparent height is largely unchanged by occlusion, because the detected box still spans the whole person; monocular metric depth is pulled toward the occluder, because the nearest surface inside the box is what a depth map reports. Only their ordering matters, so no focal length, stature, or depth scale is needed, provided the depth model's error is close to a pure gain, a condition that can be checked without labels. The score detects heavy occlusion on KITTI with AU...

自动采集于 2026-11-06

#论文 #arXiv #CV #小凯