CRUISEResearchGroup/AnyMo-Bench
收藏资源简介:
AnyMo Bench是一个具有挑战性的细粒度野外人类活动识别基准,基于Nymeria数据集中的真实可穿戴惯性测量单元流构建,为可穿戴运动识别提供了未见受试者和跨设备评估设置。该基准包含来自196名受试者的154,695个合格活动窗口,覆盖211.6小时的真实野外IMU数据。IMU流同步到60Hz的时间网格,每行包含一个活动窗口,其中imu数组的形状为[T, 18],T小于等于300,18个通道分别连接头、左手腕和右手腕各一个选定的IMU。
AnyMo Bench is a challenging fine-grained in-the-wild HAR benchmark built from real wearable IMU streams in the Nymeria dataset. It provides unseen-subject and cross-device evaluation settings for wearable motion recognition. The benchmark contains 154,695 eligible activity windows from 196 subjects, covering 211.6 hours of real in-the-wild IMU data. IMU streams are synchronized to a common 60 Hz temporal grid. Each row contains one activity window with an `imu` array of shape `[T, 18]`, where `T <= 300` and the 18 channels concatenate one selected IMU from each of Head, Left Wrist, and Right Wrist.




