IMG2IMU: Translating Knowledge from Large-Scale Images to IMU Sensing Applications

arXiv:2209.00945v2 Announce Type: replace
Abstract: Pre-training representations acquired via self-supervised learning could achieve high accuracy on even tasks with small training data. Unlike in vision and natural language processing domains, pre-training for IMU-based applications is challenging, as there are few public datasets with sufficient size and diversity to learn generalizable representations. To overcome this problem, we propose IMG2IMU that adapts pre-trained representation from large-scale images to diverse IMU sensing tasks. We convert the sensor data into visually interpretable spectrograms for …

abstract accuracy acquired and natural language processing applications arxiv cs.lg data datasets diversity domains images knowledge language language processing learn natural natural language natural language processing pre-training processing public scale self-supervised learning sensing small supervised learning tasks training training data type via vision