Daily Activity Recordings for artificial intelligence (DARai)
收藏资源简介:
DARai数据集是一个多模态、分层标注的数据集,旨在理解现实世界中的人类活动。该数据集包含了50名参与者在10个不同环境中超过200小时的连续记录,数据来自20个传感器,包括多角度摄像头、深度和雷达传感器、可穿戴惯性测量单元(IMU)、肌电图(EMG)、鞋垫压力传感器、生物监测传感器和凝视跟踪器。DARai数据集在三个层次上进行标注:高级活动(L1)、低级动作(L2)和细粒度步骤(L3),以捕捉人类活动的复杂性。数据集的标注和记录设计允许22.7%的L2动作在L1活动之间共享,14.2%的L3步骤在L2动作之间共享。DARai数据集的开放性和多传感器设计使其能够支持现实世界中人类活动理解的挑战性应用。
The DARai dataset is a multimodal, hierarchically annotated dataset designed for understanding human activities in the real world. This dataset contains over 200 hours of continuous recordings from 50 participants across 10 distinct environments, sourced from 20 sensors including multi-angle cameras, depth and radar sensors, wearable inertial measurement units (IMUs), electromyography (EMG) sensors, insole pressure sensors, biometric monitoring sensors, and gaze trackers. The DARai dataset is annotated at three levels: high-level activities (L1), low-level actions (L2), and fine-grained steps (L3), to capture the complexity of human activities. The annotation and recording designs of the dataset enable 22.7% of L2 actions to be shared across L1 activities, and 14.2% of L3 steps to be shared across L2 actions. The open-access nature and multi-sensor design of the DARai dataset enable it to support challenging applications for real-world human activity understanding.

- 1Hierarchical and Multimodal Data for Daily Activity UnderstandingOLIVES at the Center for Signal and Information Processing CSIP, School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, USA, Amazon Lab126, San Francisco, CA, USA · 2025年



