ThermoHands

ThermoHands是由爱丁堡大学创建的第一个针对热图像基的自我中心3D手势估计基准数据集。该数据集包含约96,000个同步的多光谱、多视角图像，捕捉了28名参与者在不同环境中的手-物体和手-虚拟交互。数据集通过创新的自动标注过程丰富了3D手势地面真实数据。ThermoHands旨在解决在各种条件下，如不同光照和手部穿戴情况下，有效使用自我中心热图像进行3D手势估计的问题，并比较了使用RGB、NIR和深度图像技术的性能。

ThermoHands is the first egocentric 3D gesture estimation benchmark dataset based on thermal imagery, created by the University of Edinburgh. This dataset contains approximately 96,000 synchronized multispectral, multi-view images that capture hand-object and hand-virtual interactions of 28 participants across diverse environments. The dataset is enriched with 3D gesture ground-truth data via an innovative automatic annotation pipeline. ThermoHands aims to address the challenge of effectively using egocentric thermal imagery for 3D gesture estimation under various conditions, such as varying lighting and hand-wearing scenarios, and benchmarks the performance of techniques utilizing RGB, NIR, and depth images.