Accurate masks in real time during movement #2
Description
This is a great implementation. The image segmentation XR sample works well in real time when the image/object is static (with little or no head movement). I’d like to have real-time tracking while moving my head (or moving the object), but all the attempts I’ve made (tweaking the number of layers per frame, the confidence threshold, etc.) have been unsuccessful; the result is always a mask that “lags” behind the motion and is no longer correctly positioned, then stabilizes after I stop moving for a few seconds.
Do you think this is actually achievable and I’m just approaching it the wrong way, or is it completely unrealistic given the current technical limitations ?