Researchers at Oxford University have created a new dataset that contains roughly 3,900 hours of data on human activity that researchers collected through wrist-worn accelerometers, wearable cameras, and sleep diaries from 151 people. The dataset can help train more accurate models that enable wrist-worn devices to recognize and interpret data on various activities, such as running or biking, and improves on conventional datasets that typically do not include such mixed and nuanced movements.
Image credit: Luke Chesser