SkyScan photographs and identifies every airplane that flies over one house

Luke Berndt's house sits under a busy stretch of sky, with aircraft streaming overhead day and night — most of them above 30,000 feet, too high and fast for a person to identify by eye. Flight trackers could name the planes on a map, but Berndt wanted the planes themselves: sharp photographs of every aircraft that crossed his patch of sky.

His system, SkyScan, gives the camera a sense of where to look. An RTL-SDR dongle and PiAware receiver decode the ADS-B broadcasts every aircraft transmits — position, altitude, speed, identity — and a control loop turns those coordinates into pan, tilt and zoom commands for a weatherproof PTZ camera mounted outdoors. As a plane crosses overhead, the camera tracks it smoothly and captures a photo; at full zoom the camera's field of view is only about one degree, so the ADS-B aiming is what makes the shot possible at all.

Each photograph is then matched against FAA aircraft registration data, so the stream shows not just a plane but who owns it — which is why so many names on the site turn out to be banks that lease aircraft. The camera was a Costar RISE 4260 with a built-in heater that Berndt bought on eBay for about $500, and the pipeline also used Azure Custom Vision alongside SkyScan's own machine-learning notebooks, turning the accumulating photos into a precisely labelled dataset for classifying aircraft types.

Berndt shared the result on Hacker News on 22 January 2022, where it drew 740 points and 182 comments. Commenters described it as a project they had dreamed of for years — one had started building a servo-gimbal version half a decade earlier — and the wave of attention saturated skybot.cam. The whole build was open-sourced under IQT Labs' SkyScan repository so others could point the same kind of camera at their own sky.