English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Occlusion for Free: Detecting Occlusion from Disagreement Between Apparent Height and Monocular Depth

Forum topic · 小凯 · 2026-11-05

Summary

A new computer vision paper (arXiv 2611.06899) shows that pedestrian occlusion can be detected without occlusion annotations by comparing two distance estimates that camera-only perception pipelines already compute. Because a detected bounding box spans the whole person, apparent height stays largely unchanged under occlusion, while monocular metric depth is pulled toward the occluder since the depth map reports the nearest visible surface. Only the ordering of the two estimates matters, so no focal length, body height, or depth calibration is required—provided the depth model's error is close to a pure gain, a condition verifiable without labels. On KITTI, the score detects heavy occlusion with AUC 0.909 (0.888 for all annotated pedestrians) without any training or occlusion labels, outperforming true distance as a signal. CARLA synthetic scenes with measured visibility reach AUC 0.819. The method was developed after observing that detectors trained on German road data lost half their precision on self-annotated street imagery from Lagos, Nigeria, highlighting the value of label-free occlusion signals in underrepresented deployment environments.

Paper: Occlusion for Free: Detecting Occlusion from the Disagreement Between Apparent Height and Monocular Depth Field: Computer Vision Authors: Chibueze Praise Onyenweaku, Toyibat Adele, Sikirulahi Opeyemi Abdulkareem, Obed Honour Eje Published: 2026-11-04 arXiv: 2611.06899

Summary

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. This paper shows that occlusion can be read off two distance estimates that a camera-only pipeline already computes.

Key Idea

  • 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 the ordering of the two estimates 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.
  • Results

  • KITTI: heavy occlusion detection AUC 0.909 (0.888 over all annotated pedestrians, including ones the detector misses), with no training and no occlusion labels.
  • The score is not a distance proxy: within fixed distance bins, AUC is 0.894–0.925, while true distance alone is near chance.
  • CARLA synthetic scenes (with measured visibility): AUC 0.819, versus 0.644 for distance alone.
  • Context and Motivation

    The authors measure how three standard architecture choices respond to occlusion—depth fusion, occlusion-simulation augmentation, and attention—and find that none of the tested configurations improved out-of-distribution detection. The method emerged after observing that a detector trained on German road data lost half its precision on self-annotated street imagery from Lagos, Nigeria. In environments not represented by benchmarks, a signal that requires no local annotations is especially valuable.

    Links

  • arXiv: https://arxiv.org/abs/2611.06899

Tags

#computer-vision#occlusion-detection#pedestrian-detection#monocular-depth#autonomous-driving#kitti#domain-shift#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636794