Inertia-1 — A Universal Motion Model
Placement transfer<br>From one wrist to the whole body Pretrained on the wrist alone, it transfers to placements it never saw — hip, ankle, chest, and more.
Multi-stream<br>One sensor becomes many Fuse extra sensors and placements and accuracy climbs — the streams are complementary, not redundant.
Rate-robust<br>Steady from 1 Hz to 20 Hz Pretrained representations stay strong even at low sampling rates, with finer rates helping subtle health signals.
Task-universal<br>One model, every task The same backbone powers activity recognition, gait analysis, and long-horizon disease prediction.
Health insight<br>Movement, read as health Passive motion carries long-horizon health markers, linking everyday movement to clinical outcomes.
Device-agnostic<br>Works across devices & sensors Robust to new devices and sensor modalities, so it plugs into whatever a wearable already has.
Inertia-1
An Open Exploration to a Unified Motion Foundation Model
Contact Us
The big idea<br>Towards one general motion model
Motion is universal — but the models built for it weren't. Inertia-1 brings the whole landscape under one roof.
01A fragmented field<br>Datasets disagree on the basics — sampling rate, window length, sensor modality, body placement, even signal format — and every task gets its own bespoke model. Findings rarely carry from one setup to the next.
02One unified exploration<br>Inertia-1 studies the full lifecycle of motion models — data, sensing, objectives, and scale — inside a single, controlled space instead of isolated one-offs.
03A general representation<br>The payoff: one representation that adapts across placements, devices, and tasks — the same backbone, working far beyond the setting it was trained on.
Head<br>Chest<br>Back<br>Arm<br>Wrist<br>Hand<br>Hip<br>Thigh<br>Knee<br>Shin<br>Ankle<br>Accelerometer<br>Gyroscope<br>Magnetometer<br>Triaxial<br>ENMO<br>0.2 Hz<br>1 Hz<br>5 Hz<br>20 Hz<br>10 s window<br>30 s window<br>60 s window<br>2 hr window<br>Frequency domain<br>Time domain<br>Activity recognition<br>Gait detection<br>Longitudinal health<br>Head<br>Chest<br>Back<br>Arm<br>Wrist<br>Hand<br>Hip<br>Thigh<br>Knee<br>Shin<br>Ankle<br>Accelerometer<br>Gyroscope<br>Magnetometer<br>Triaxial<br>ENMO<br>0.2 Hz<br>1 Hz<br>5 Hz<br>20 Hz<br>10 s window<br>30 s window<br>60 s window<br>2 hr window<br>Frequency domain<br>Time domain<br>Activity recognition<br>Gait detection<br>Longitudinal health
body) ============ -->
What we found<br>Beyond benchmarks, Inertia-1 surfaces the choices that decide whether a motion model actually works in the real world.
Learn it on the wrist. Use it anywhere on the body.
Discover more
Go back<br>Pretrain once on the wrist, then point the model anywhere. It holds up on body placements — and even sensor types like gyroscope and magnetometer — that it never saw during training. No retraining for each new spot on the body.
Wrist accelerometer<br>Other sensors · gyro, mag<br>Other placements
Fused representation<br>higher accuracy · cleaner motion clusters
Add more streams. Get more signal.
Discover more
Go back<br>Stack on more streams — extra placements, gyroscope, magnetometer — and the learned representation gets both more accurate and cleaner, with activities separating into tighter clusters. The streams are complementary: each one catches something the others miss.
Also worth knowing<br>Sensing design is a first-order choice
How you capture motion shapes what a model can do with it. A few practical rules of thumb from the study.
Sampling rate<br>Pretrained models stay strong even at a low 1 Hz for activity recognition; finer-grained health signals benefit from higher sampling rates.
Window length<br>30–60 second windows hit the sweet spot across most tasks — long enough to capture context, short enough to stay sharp.
Keep all three axes<br>Full triaxial input consistently beats collapsed vector-magnitude summaries — the extra axes carry signal worth keeping.
Stay in the time domain<br>Time-domain modeling preserves gait and health cues better than frequency-domain reconstruction.
How it works<br>One pipeline, from raw signal to real-world insight
The general representation comes together in three clean steps.
01
Pretrain at scale
Learn from planetary-scale accelerometry — over 18 million hours across global cohorts — with self-supervision, no labels required.
02
Transfer across settings
Adapt the same representation to new placements, devices, and sampling rates with light tuning — or none at all.
03
Deploy across tasks
Power activity, mobility, and health applications from one backbone — from fitness tracking to clinical screening.
Capabilities<br>From movement to meaning
The same representation spans the full spectrum of motion understanding.
Activity & behavior
Recognize everyday activities and behavioral patterns with state-of-the-art accuracy across diverse populations.
Gait & mobility
Detect subtle gait changes — like freezing of gait — that signal mobility decline and neurological conditions.
Health & disease
Surface long-horizon health markers from passive motion, linking everyday movement to clinical...