This repo accompanies the research paper, ARKitScenes - A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile RGB-D Data and contains the data, scripts to visualize and process assets, and training code described in our paper.
ARKitScenes - A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile RGB-D Data
upon using these data or source code, please cite
@inproceedings{
dehghan2021arkitscenes,
title={{ARK}itScenes - A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile {RGB}-D Data},
author={Gilad Baruch and Zhuoyuan Chen and Afshin Dehghan and Tal Dimry and Yuri Feigin and Peter Fu and Thomas Gebauer and Brandon Joffe and Daniel Kurz and Arik Schwartz and Elad Shulman},
booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1)},
year={2021},
url={https://openreview.net/forum?id=tjZjv_qh_CE}
}
ARKitScenes is not only the first RGB-D dataset that is captured with now widely available depth sensor, but also is the largest indoor scene understanding data ever collected. In addition to the raw and processed data, ARKitScenes includes high resolution depth maps captured using a stationary laser scanner, as well as manually labeled 3D oriented bounding boxes for a large taxonomy of furniture. We further provide helper scripts for two downstream tasks: 3D object detection and RGB-D guided upsampling. We hope that our dataset can help push the boundaries of existing state-of-the-art methods and introduce new challenges that better represent real world scenarios.
• ARKitScenes is the first RGB-D dataset captured with the widely available Apple LiDAR scanner. Along with the raw data we provide the camera pose and surface reconstruction for each scene.
• ARKitScenes is the largest indoor 3D dataset consisting of 5,047 captures of 1,661 unique scenes.
• We provide high quality ground truth of (a) registered RGB-D frames and (b) oriented bounding boxes of room defining objects.
Below is an overview of RGB-D datasets and their ground truth assets compared with ARKitScenes. HR and LR represent High Resolution and Low Resolution respectively, and are available for a subset of 2,257 captures of 841 unique scenes.
In the figure below, we provide (a) illustration of iPad Pro scanning set up. (b) mesh overlay to assist data collection with iPad Pro. (c) example of one of the scan patterns captured with the iPad pro, the red markers show the chosen locations of the stationary laser scanner in that room.
To download the data please follow the data documentation
Here we provide the two tasks mentioned in our paper, namely, 3D Object Detection (3DOD) and depth upsampling.
Please refer to the LICENSE file for detailed information on using the dataset.
For any additional inquiries about the license, please contact [email protected].
If you have other questions, feel free to open an issue in the repository or contact [email protected].