🧵#Positioning in AR

Generally that's quite a challenging problem and the solution differs a bit regarding size/distance of the virtual object and also what control you have over devices.

The highest accuracy is usually with image tracking – which works out of the box on iOS but currently requires users to enable a browser flag on Android (so it's feasible when you have some control over the devices but not in a general case)

Lower accuracy is using geolocation (accurate to 2-5m, depending on where you are in the world), which is suitable for large-scale setups (e.g. virtual building should show up in the place where it will be built)

If you have a controlled environment (e.g. a known space that you can scan upfront) then using camera data and doing some visual mapping is the best approach (this is similar to what Niantic's or Google's VPS is doing) but also technically the most challenging.

Seems like the last option is the right pick. Can you please point me in where to start?