[Feature Request] Machine learning for tracking algorithms #291

Closed
opened 2026-04-05 18:17:42 +02:00 by MrUnknownDE · 0 comments
Owner

Originally created by @Znunu on 7/30/2025

Because of the inherent nature of IMU tracking, slimes have lots of issues with tracking. From my understanding, various algorithms exist in the code that attempt to fix this (e.g. the recent stay aligned update, the various calibration poses etc). Yet none fix the issue entirely and they all rely on manual tuning of various parameters.

To me, this sounds like a problem where ML would be worth trying. Especially with so much training data available, from many people using slimeVR. Even without a true reference available for this data, it might still be possible to figure something out.

*Originally created by @Znunu on 7/30/2025* Because of the inherent nature of IMU tracking, slimes have lots of issues with tracking. From my understanding, various algorithms exist in the code that attempt to fix this (e.g. the recent stay aligned update, the various calibration poses etc). Yet none fix the issue entirely and they all rely on manual tuning of various parameters. To me, this sounds like a problem where ML would be worth trying. Especially with so much training data available, from many people using slimeVR. Even without a true reference available for this data, it might still be possible to figure something out.
Sign in to join this conversation.
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github/SlimeVR-Server#291