Ghost Detector Lab

What is an SLS camera?

A structured light sensor camera. The originals were built out of a games console depth camera: a projector threw a grid of infrared dots across a room, a sensor watched how that grid deformed, and software looked through the result for something human-shaped.

Does this SLS camera actually detect ghosts?

No. Nothing on this site detects ghosts and no instrument ever demonstrated has. This instrument does two separate things and the panel counts them separately: it runs a real pose estimation model on your camera feed, which finds real people, and it draws its own figures into regions of the frame it cannot resolve. The second kind are drawn by the instrument.

What is the difference between RETURN and RESOLVE on the panel?

RETURN counts tracked landmark sets — figures a pose estimation model found in the actual video. RESOLVE counts reconstructions, which are figures this instrument drew. They are separate readouts, they are stored as separate numbers, and nothing on the site ever adds them together.

What is the resolution floor?

A real measurement of your camera's frame. The array divides the picture into 192 tiles and scores each one for how dark it is, how much it changes between frames relative to its brightness, and how little structure it contains. Those three things are why pose models fail into false positives, so the score is an honest answer to the question of where the array cannot tell what it is looking at.

Why do figures only appear in dark corners?

Because that is where the instrument is allowed to draw them. A reconstruction resolves out of the highest-scoring tile of the resolution floor, and the rate it arrives at is a function of that tile's score. In a brightly lit room the score never crosses the working gate and no figure will ever appear, however long you wait. This is deliberate and it is the closest thing the instrument has to an honest rule.