HPointLoc

该数据集名为HPointLoc，专为室内环境下的视觉地标识别以及同时定位与地图构建（SLAM）中的循环检测而设计。它包含了来自Matterport3D的49个场景。数据集不仅提供了RGB和深度图像，还包括了41个类别的实例分割以及带有指南针数据的GPS信息。该数据集被划分为验证集（HPointLoc-Val）和完整集（HPointLoc-All）。其规模涵盖了49个场景，任务重点在于视觉地标识别和循环检测。

This dataset, named HPointLoc, is specifically designed for visual landmark recognition in indoor environments and loop closure detection in Simultaneous Localization and Mapping (SLAM). It contains 49 scenes sourced from the Matterport3D dataset. In addition to RGB and depth images, the dataset also provides instance segmentation masks for 41 categories, as well as GPS data along with compass data. The dataset is split into two subsets: the validation set (HPointLoc-Val) and the full set (HPointLoc-All). Covering 49 scenes in total, the dataset focuses on the core tasks of visual landmark recognition and loop closure detection.