Occlusion detection from disagreement between two monocular distance cues
Original title: Occlusion for Free: Detecting Occlusion from the Disagreement Between Two Monocular Distance Cues
Occlusion read from apparent height and monocular metric depth disagreement, no training or labels required.
Occlusion can be detected using apparent height and monocular metric depth from a camera-only pipeline. The score requires no focal length or stature and works when depth error is close to a pure gain. It detects heavy occlusion on KITTI with AUC 0.909 and reaches 0.819 on synthetic CARLA scenes.
- No occlusion annotations or training needed
- Apparent height largely unchanged by occlusion
- Monocular metric depth pulled toward occluder
- AUC 0.909 on KITTI, 0.819 on CARLA
- Standard architectural responses to occlusion do not help
From the original article
1 min read at arxiv.orgKnowing 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.
By Chibueze Praise Onyenweaku · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.