NAVER LABS localization datasets
收藏资源简介:
NAVER LABS localization datasets是首个提出大规模密集采样室内环境的数据集,包含动态物体、移动人群和变化场景等挑战。数据集在韩国首尔的大型购物中心和地铁站通过配备激光扫描器、摄像头和轮式里程计的专用映射平台捕获。数据集创建过程中,开发了一种基于激光雷达同步定位与地图构建(SLAM)和运动恢复结构(SFM)的新管道,以生成地面实况姿态。这些数据集旨在推动视觉定位、图像检索和局部特征提取等领域的研究,并可能用于深度估计、激光雷达地点识别和地图变化检测等任务。
NAVER LABS Localization Datasets is the first dataset focused on large-scale densely sampled indoor environments, which features challenges such as dynamic objects, moving crowds, and changing scenes. The datasets were captured in large shopping malls and subway stations in Seoul, South Korea, using a dedicated mapping platform equipped with laser scanners, cameras, and wheel odometry. During the dataset creation process, a novel pipeline based on LiDAR Simultaneous Localization and Mapping (SLAM) and Structure from Motion (SfM) was developed to generate ground-truth poses. These datasets aim to advance research in fields including visual localization, image retrieval, and local feature extraction, and can also be applied to tasks such as depth estimation, LiDAR place recognition, and map change detection.

- 1Large-scale Localization Datasets in Crowded Indoor SpacesNAVER LABS · 2021年



