Event-based Segmentation Dataset (ESD)

Event-based Segmentation Dataset (ESD) 是由先进研究与创新中心开发的，专为室内杂乱环境中的物体分割任务设计的高质量3D时空数据集。该数据集包含145个序列，共计14,166个手动标注的RGB帧，以及来自两个事件相机的总计42.68百万事件。ESD数据集通过提供密集标注和3D时空事件基准，旨在为社区提供一个具有挑战性的分割基准。数据集的创建过程涉及使用UR10机器人和多种相机配置，以捕捉不同条件下的数据。ESD数据集的应用领域包括机器人感知、物体定位和分割，特别是在非结构化环境中，旨在提高机器人在复杂环境中的操作能力。

Event-based Segmentation Dataset (ESD) is a high-quality 3D spatiotemporal dataset developed by the Advanced Research and Innovation Center, specifically designed for object segmentation tasks in cluttered indoor environments. This dataset includes 145 sequences, a total of 14,166 manually annotated RGB frames, and a total of 42.68 million events collected from two event cameras. The ESD dataset aims to provide the research community with a challenging segmentation benchmark by offering dense annotations and 3D spatiotemporal event benchmarks. The creation of the ESD dataset employed the UR10 robot and multiple camera configurations to capture data under various conditions. Application scenarios of the ESD dataset include robotic perception, object localization and segmentation, especially in unstructured environments, with the aim of improving robots' operational capabilities in complex environments.