LIDAR-A2S Dataset (v1.0): A Multi-View Ankle-Level 2D LiDAR Dataset with Synchronized 3D Skeleton Annotations for Privacy-Preserving Human Motion Analysis

LiDAR-A2S is a multi-view, ankle-level 2D LiDAR dataset with synchronized 3D human skeleton annotations, designed for research on multi-object tracking, human pose estimation, and privacy-preserving human sensing in indoor environments. The dataset was collected using four Hokuyo URG 2D LiDAR sensors positioned at ankle height and a Kinect v2 sensor used solely for internal validation and ground-truth skeleton extraction. Raw RGB images captured by the Kinect sensor are not included in the public release due to privacy considerations. The publicly available data consist of synchronized 2D LiDAR scans, derived LiDAR projection images, and per-frame 3D skeleton joint coordinates. Data were recorded in controlled indoor environments under varying human motion patterns, including walking, turning, stopping, and interactions among multiple individuals. All sensor streams were temporally synchronized, and calibration procedures were applied to ensure spatial alignment across sensors. This dataset is intended to support benchmarking and development of algorithms for multi-object tracking, skeleton mapping from sparse range data, and privacy-awarehuman behaviour analysis. Detailed descriptions of the data formats, coordinate systems, and preprocessing steps are provided in the accompanying README file.