Variable Input Masked Autoencoders for Wrist-Worn Accelerometry
Large-scale repositories of wearable sensor data have enabled the development of foundation models for human activities and behavior. However, existing models are pretrained using fixed input durations and sampling rates, despite substantial heterogeneity in recording conditions, downstream tasks, and deployment constr...
H. Haresamudram, Wan-Ting Mao, Aditya Radhakrishnan et al.
· 0 citations