CityRefer

CityRefer数据集是由ATR、东京工业大学和RIKEN AIP共同创建的，旨在推动城市级3D视觉定位技术的研究。该数据集包含超过35,000个自然语言描述，用于在城市规模的点云数据中定位3D对象。数据集中的描述与OpenStreetMap中的5,000个地标标签同步，所有数据均经过人工验证，确保高质量和准确性。CityRefer数据集的应用领域包括自动驾驶车辆的交互式导航和无人机导航，旨在解决城市环境中3D对象的精确定位问题。The CityRefer dataset was co-created by ATR, Tokyo Institute of Technology, and RIKEN AIP, aiming to advance research on urban-scale 3D visual localization technologies. This dataset contains over 35,000 natural language descriptions for localizing 3D objects in city-scale point cloud data. The descriptions in the dataset are synchronized with 5,000 landmark tags from OpenStreetMap, and all data has been manually verified to ensure high quality and accuracy. The application scenarios of the CityRefer dataset include interactive navigation for autonomous vehicles and drone navigation, which aims to solve the problem of precise 3D object localization in urban environments.

搜集汇总

背景与挑战

背景概述

CityRefer数据集由ATR、东京工业大学和RIKEN AIP共同创建，旨在推动城市级3D视觉定位技术研究。它包含超过35,000个自然语言描述，与OpenStreetMap中的5,000个地标标签同步，并经过人工验证，确保高质量和准确性，主要应用于自动驾驶车辆和无人机导航，以解决城市环境中3D对象的精确定位问题。