ThermoHands

该数据集是首个专注于基于热成像自拍摄三维手部姿态估计的基准，涵盖了多视角和多光谱的数据集。该数据集由28位受试者在多种情境下进行手物互动和手虚拟互动时收集而成。数据集通过自动化的流程进行了三维手部姿态的标注，并包含了基于关节误差的准确性评估。该数据集规模涉及28位受试者，其任务是进行三维手部姿态估计。

This dataset is the first benchmark dedicated to self-captured thermal imaging-based 3D hand pose estimation, covering multi-view and multi-spectral data. It was collected from 28 subjects performing hand-object interactions and hand-virtual interactions in various scenarios. The 3D hand pose annotations of this dataset are generated via an automated pipeline, and it also includes accuracy evaluations based on joint error metrics. With data from 28 subjects, this benchmark focuses on the task of 3D hand pose estimation.