niessner / matterport

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

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.