Hold the camera level. Place a measured vertical reference centered in the full image, perpendicular to the optical axis. Measure its axial distance from the camera lens (not a slanted tape distance). Freeze the image, then mark its top and bottom. Do not change lens, zoom, resolution or height afterward.

Tap both reference endpoints above, or enter their positions below (0% = image top). The reference must be vertically centered.

Tap the live depth image to place the crosshair at a visible ground-contact point. This is not an automatic distance to a detected object. Keep the selected point on flat, level ground.

Intervals describe optical-axis depth, not nearest-obstacle clearance. Bounds assume ±2° motion error and measured tolerances, not a validated confidence probability. Stop and recalibrate after any camera-height or projection change.

Optional download: approximately 50 MB on first use (model and runtime). Geometry works without it. Downloads may fail offline.

Depth unavailable. Optional model off.

Snapshot, not live. Relative depth only.

Historical snapshot only, not live depth or metres. Analysis may take several seconds. The frozen source image and map remain paired until cleared. No same-frame metric reference fit is applied.