交通基础设施多模态三维构件数据集

随着传统交通基础设施建管养模式的智能化重构，三维对象语义解析能力缺失问题日益突出。为系统性破解行业智能化升级的三维数据匮乏瓶颈，中国交通建设集团有限公司联合中国交通信息科技集团有限公司，通过源数据采集、数据处理、数据标注，数据建模、数据应用、以及数据质量控制和模型闭环迭代的“5+2”建设架构，共同打造了约11.8TB 的交通基础设施多模态三维构件数据集，构建近6万个高质量三维模型、覆盖 2327个精细分类、包含点云图像文本等多种模态标注数据，精准填补了全球交通基建领域三维多模态数据集空白。With the intelligent restructuring of the construction, management and maintenance models of traditional transportation infrastructure, the problem of lacking semantic parsing capabilities for 3D objects has become increasingly prominent. To systematically solve the bottleneck of 3D data shortage in the industry's intelligent upgrading, China Communications Construction Group Co., Ltd. and China Transportation Information Technology Group Co., Ltd. jointly developed a multi-modal 3D component dataset for transportation infrastructure with a total scale of approximately 11.8 TB, based on the "5+2" construction framework covering source data collection, data processing, data annotation, data modeling, data application, as well as data quality control and closed-loop model iteration. The dataset encompasses nearly 60,000 high-quality 3D models, covers 2,327 fine-grained categories, and includes multi-modal annotated data such as point clouds, images and texts, which accurately fills the global gap in 3D multi-modal datasets for the transportation infrastructure sector.

搜集汇总

背景与挑战

背景概述

该数据集由中国交通建设集团与信科集团联合构建，规模约11.8TB，包含近6万个高质量三维模型，覆盖2327个精细分类，融合了点云、图像和文本三种模态数据。它旨在填补交通基础设施领域三维多模态数据集的空白，支持智能设计、车路协同感知和数字孪生运维等前沿应用。