Measuring the world with a tap

DepthTag runs an ARKit world-tracking session with one extra frame semantic switched on: sceneDepth. Every frame then carries an ARDepthData whose depthMap is a buffer of Float32 distances, in meters, straight from the LiDAR sensor.

When you tap, the screen point is converted to a coordinate in that depth buffer and the distance is read directly — no photogrammetry, no guessing, just the sensor's own measurement. The tag is then placed in world space so it stays put as you move around it.

The important engineering decision is the negative space: the entire feature is gated behind a LiDAR capability check. On a phone without the sensor there is no honest depth to report, so DepthTag doesn't pretend. A measuring tool that lies is worse than one that says "not on this device."