Matterport

niessnerMIT长期未更新

Matterport3D 是一个面向 RGB-D 机器学习任务的大规模室内场景数据集，提供带标注的 3D 重建与语义理解数据，适合训练和评测场景重建模型。

Matterport3D is a pretty awesome dataset for RGB-D machine learning tasks :)

暂无版本1,269star155fork同步于 2026-11-07 07:30 UTC

RGB-D数据集 3D重建语义场景理解室内场景深度学习

对比 GitHub 官网 3d-reconstruction semantic-scene-understanding

怎么上手

由 README 提炼，执行前请确认

原文

README 未提供安装命令，数据需签署使用协议后申请下载

硬件要求按 README 说明

README 未说明

下载：源码 zip·GitHub Releases（安装包）本站不镜像安装包

项目说明

由 README 自动翻译提炼

没人工点评？写下你的经验

Matterport3D 是面向 RGB-D 室内场景理解的大规模数据集，附带标注、基准任务与加载工具

Matterport3D 是一个面向 RGB-D 机器学习任务的大规模室内场景数据集，提供带标注的 3D 重建与语义理解数据，适合训练和评测场景重建模型。

核心能力

包含 90 个房产的彩色与深度图像、相机位姿
提供带纹理 3D 网格、楼层平面图与区域标注
提供物体实例语义标注
定义关键点匹配、视角重叠预测等基准任务
附带数据加载与可视化代码

容易踩的坑

数据集不能直接下载，需用机构邮箱签署 Terms of Use 并发送邮件申请访问权限

内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README

用户点评

登录用户的真实使用记录，可按版本和硬件筛选

共 0 条

还没有人点评。第一个说出真实体验的人最有价值。

README 原文

未翻译

在 GitHub 打开

Matterport3D

Matterport3d

The Matterport3D V1.0 dataset contains data captured throughout 90 properties with a Matterport Pro Camera.

This repository includes the raw data for the dataset plus derived data, annotated data, and scripts/models for several scene understanding tasks.

Visit the main website for updates and to browse the data.

Paper

Matterport3D: Learning from RGB-D Data in Indoor Environments

If you use the Matterport3D data or code please cite:

@article{Matterport3D,
  title={{Matterport3D}: Learning from {RGB-D} Data in Indoor Environments},
  author={Chang, Angel and Dai, Angela and Funkhouser, Thomas and Halber, Maciej and Niessner, Matthias and Savva, Manolis and Song, Shuran and Zeng, Andy and Zhang, Yinda},
  journal={International Conference on 3D Vision (3DV)},
  year={2017}
}

Data

The dataset consists of several types of annotations: color and depth images, camera poses, textured 3D meshes, building floor plans and region annotations, object instance semantic annotations. For details see the data organization document.

To download the dataset, you must indicate that you agree to the terms of use by signing the Terms of Use agreement form, using your institutional email addresses, and sending it to: [email protected]. We will then provide download access to the dataset.

Benchmark tasks

Using the Matterport3D data, we present several benchmark tasks: image keypoint matching, view overlap prediction, surface normal estimation, region type classification, and semantic voxel labeling. See the tasks directory for details.

Tools

We provide code for loading and viewing the data. See the code directory for details.

License

The data is released under the Matterport3D Terms of Use, and the code is released under the MIT license.