HUMAR-2024
收藏资源简介:
HUMAR-2024数据集是由LIRMM, University of Montpellier, Montpellier, France等机构创建的,包含11名参与者进行的各种动作。数据集包含2200个动作实例,涵盖了17个不同的动作标签,包括站立、行走、坐下、蹲下等。数据集旨在用于实时人体动作识别,通过使用双全局快门高清灰度相机捕获视频数据,并使用双目视觉技术进行3D姿态估计。数据集的创建过程严格遵循了国际生物力学协会的指南,确保了生物力学的准确性和一致性。该数据集适用于人机交互、协作机器人等领域,旨在解决实时人体动作识别的挑战。
The HUMAR-2024 dataset was developed by institutions including LIRMM, University of Montpellier, Montpellier, France, and other relevant entities. It includes 2200 action instances collected from 11 participants, covering 17 distinct action labels such as standing, walking, sitting, squatting, and more. Designed for real-time human action recognition, this dataset captures video data via dual global-shutter high-definition grayscale cameras and performs 3D pose estimation using binocular vision technology. The dataset was constructed strictly in compliance with the guidelines of the International Society of Biomechanics, ensuring biomechanical accuracy and consistency. It is applicable to domains such as human-computer interaction and collaborative robotics, aiming to resolve the challenges associated with real-time human action recognition.

- 1Biomechanically consistent real-time action recognition for human-robot interactionLIRMM, University of Montpellier, Montpellier, France; LAAS-CNRS, Université Paul Sabatier, CNRS, Toulouse, France; Nantes Université, École Centrale Nantes, CNRS, LS2N, UMR 6004, 1, rue de la Noe, 44321 Nantes, France; IPAL, CNRS · 2025年



