LasHeR

LasHeR由1224个可见光和热红外视频对组成，总共有超过730K帧对。每个帧对在空间上对齐，并使用边界框手动注释，从而使数据集得到良好且密集的注释。LasHeR是从广泛的对象类别，相机视点，场景复杂性和环境因素中捕获的高度多样化的季节，天气，白天和黑夜。在实际应用程序的诱导下，数据创建中考虑了几个新的挑战。

LasHeR comprises 1,224 visible and thermal infrared video pairs, with a total of over 730,000 frame pairs. Each frame pair is spatially aligned and manually annotated with bounding boxes, resulting in high-quality and dense annotations for the dataset. The dataset features high diversity across seasons, weather conditions, day and night periods, and is collected from a wide range of object categories, camera viewpoints, scene complexities and environmental factors. Several novel challenges have been incorporated into the dataset creation, motivated by real-world applications.

AI搜集汇总

背景与挑战

背景概述

LasHeR是一个大规模、高多样性的RGB-T（可见光与热红外）跟踪基准数据集，包含1224个空间对齐的视频对和超过73万帧对，每帧都经过密集的边界框手动注释。数据集覆盖了广泛的对象类别、相机视点、场景复杂性和环境条件（如季节、天气和昼夜变化），旨在支持实际应用中的多模态目标跟踪研究。