Hello, I’m attempting to get head tracking data using the latest LeiaSR SDK and have tried to use the headposetracker/headposelistener, but the data from them is too noisy to utilize. For example, pitch will jitter all over the place a few degrees and then a huge 20 degree jump. I’ve tried to implement some filters, but it seems like it’s not possible to get rid of all the jitter and the stronger the filter, the less responsive head tracking is. Am I missing something? or is there a better way to get head tracking data?
1 Like
I’m not sure about the code, but as far as I remember the headtracking was smooth, so you might be doing something wrong. Note, with head-tracking in general (for XR), you need to use sensor fusion, the raw data is always unusable. You can either do this manually (e.g. with a madgwick filter) or use a different API, I imagine the SDK would do this for you but I’m not sure.
Hmm… in the LeiaSR Tracking app, I do see the same jitter that I also get in test code.
I also tried to fuse between predictingeyetracker and the headposetracker, but didn’t have much luck there. What could I sensor fuse with? I’m not looking to add an IMU to my head haha.
Interestingly, running the simple OpenTrack NeuralNet tracking using the same LeiaSR camera is a lot smoother.