Model

PRIMUS

Nokia Bell Labs pretrained IMU encoders (Clear BSD)

Cite
UnconfirmedCatalog paper date: 2025Source

Activity counts measure attention, not model quality. Missing counts are unknown. Publication status follows the linked source; a listed venue alone does not confirm peer review.

PRIMUS is a pre-training method for IMU motion-sensor representations using multimodal and self-supervised learning (arXiv 2411.15127; oral at ICASSP 2025). The official Nokia Bell Labs repository releases training code, and a pretrained checkpoint is on Zenodo (10.5281/zenodo.15147513).

Form factor
wearable_IMU
Signals
IMU
Year
2025
License

Clear BSD (BSD-3-Clause-Clear) on Nokia-Bell-Labs/pretrained-imu-encoders LICENSE

At a glance
Type
Task model
Modalities
IMU
Form factor
Wearable IMU
Parameters
Not reported
Architecture
Not reported
Venue
Not reported
Year
2025
arXiv
Not reported
Code
Not reported
Weights
Open
Licence
Clear BSD (BSD-3-Clause-Clear) on Nokia-Bell-Labs/pretrained-imu-encoders LICENSE
Caveats
Not reported
Datasets
Not reported
Added
Not reported
Updated
Nov 9, 2026
PRIMUS: ChatPPG research summary card

How it works

How PRIMUS works: Input signals → Model → Downstream tasksInput signals: IMU (Accelerometer, 3-axis). Model: PRIMUS.Input signalsIMU: Accelerometer, 3-axisxyzIMUAccelerometer, 3-axisModelPRIMUS
Inputs
IMU
Approach
IMU · Nokia · BSD · teacher candidate

Datasets used

No public datasets linked yet.

Cite this model

Cite the publication recorded for this model. Check it against the publisher before submitting.

PRIMUS. (2025). https://doi.org/10.5281/zenodo.15147513
@misc{primusimu2025primus,
  title         = {{PRIMUS}},
  year          = {2025},
  doi           = {10.5281/zenodo.15147513}
}

History

  1. Record updated