LSM-2: Learning from Incomplete Wearable Sensor Data
Paper • 2506.05321 • Published
Reference checkpoint for OpenMHC. Window: daily (1440 minute, patch 10), 19 sensor channels.
An adaptation of Google's LSM2 wearable foundation model to OpenMHC. LSM2 is a masked-autoencoder ViT for 1-D wearable signals; this checkpoint was pre-trained self-supervised on daily (1440 minute) OpenMHC windows and serves as a general-purpose encoder. The wrapper here exposes it for the OpenMHC imputation evaluation track.
from openmhc.imputers import LSM2Imputer
import openmhc
imp = LSM2Imputer.from_release("hf://MyHeartCounts/openmhc-lsm2-daily")
results = openmhc.evaluate_imputation(imp, version="xs")
print(results.summary())
Requires the matching optional extras: pip install 'openmhc[lsm2,hf]'.
Pin a specific revision with the @ suffix:
LSM2Imputer.from_release("hf://MyHeartCounts/openmhc-lsm2-daily@v1.0")
MHC_Dataset/mhc-mae-ssl-daily/mae-daily:v0openmhc_manifest.json.Released under the OpenRAIL license. See the OpenMHC repository for use restrictions tied to the underlying data agreement.
@misc{openmhc,
title = {OpenMHC: Accelerating the Science of Wearable Foundation Models},
author = {OpenMHC team},
url = {https://github.com/AshleyLab/myheartcounts-dataset}
}