Dear Pupil Labs Team,

My name is Ioannis Panagiotopoulos and I’m a Master’s student at TU Wien. I am currently working on my master's thesis, which focuses on estimating head roll, pitch, and yaw purely from the RGB world camera of Pupil Invisible, and I use the built-in IMU mainly for validation.

I have a clarification question regarding the IMU orientation data exported from Pupil Cloud:

According to the documentation, roll and pitch are computed using Madgwick’s algorithm based on inertial measurements. When exporting roll and pitch from Pupil Invisible recordings, do these values represent the raw gravity-referenced orientation of the headset IMU, or are they in any way compensated/corrected for whole-body motion (e.g. walking or gait-induced movement)?

This distinction is important for how I interpret and describe the IMU signals in my thesis methodology.

Thanks a lot in advance for any clarification!

To be more specific, what I am trying to understand is whether the reported roll, pitch values change purely due to the wearer’s translation through space, even if the head itself remains rigidly oriented.

For example, if a person is walking straight ahead while intentionally keeping their head orientation fixed (i.e. no deliberate head rotation), would the exported roll/pitch remain stable, or would they still exhibit variations induced by gait-related motion (e.g. vertical oscillations, body sway)?

In other words, are these orientation values absolute (gravity-referenced, world-aligned) outputs of the IMU sensor fusion, or are they effectively relative and influenced by whole-body locomotion, without any explicit compensation for walking dynamics?

This distinction is important for determining whether the signals can be interpreted as head orientation in a global frame, or whether additional modeling of gait-related effects is required on my side.

Hope this clarifies what I meant.

Thanks a lot for the clear explanation — that answers my question perfectly. This distinction is exactly what I needed for interpreting the IMU signals correctly in my analysis.

Much appreciated!