Occlusion for Free: Detecting Occlusion from the Disagreement Between Two Monocular Distance Cues

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 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. Because only their ordering matters, calibration is unnecessary. It detects heavy occlusion on KITTI with AUC 0.909, requires no training and no occlusion labels, and is not a proxy for range, since within fixed distance bands it holds AUC 0.894--0.925 while true distance alone sits at chance. We arrived here after measuring the three standard architectural responses to occlusion, depth fusion, occlusion-simulating augmentation, and attention, none of which improved detection outside its training distribution, and after finding that transferring a detector from German roads to self-annotated street scenes in Lagos, Nigeria costs half its accuracy while matching training density recovers it. Signals that need no local annotation are worth more in settings that benchmarks do not represent.