Realsee3D is a large-scale, multi-view RGB-D dataset containing 10,000 indoor scenes, comprising real-world residential scenes captured by 3D LiDAR Camera and procedurally generated scenes.
- Large Scale: 10,000 unique indoor scenes, comprising 95,962 rooms and 299,073 viewpoints/RGB-D pairs.
- Rich Data: Panoramic RGB-D captures with complete room-level coverage.
- Comprehensive Annotations: Includes pixel-wise semantic segmentation (available for both real-world and synthetic scenes), with CAD drawings, floor plans, 3D detection labels and more (forthcoming).
- Diverse Scenes: Comprising 1,000 real-world scenes with varied layouts and decoration styles, and 9,000 procedurally generated scenes utilizing over 100 designer-curated style templates, ensuring diverse furniture models and styles for robust training and testing.
The Realsee3D dataset is organized into individual scenes, each containing detailed multi-view RGB-D data.
- Data Structure & Usage: DATASET_STRUCTURE.md provides a comprehensive explanation of the file organization, data formats, and detailed usage instructions.
- Methodology: Visit our official website for more detailed information on how real-world data is collected and how synthetic data is generated.
To access the dataset, you must sign a Data Usage Agreement (PDF format). You can download the agreement directly here. Please send your request, specifying your intended use, to developer@realsee.com with the subject [Realsee3D Dataset Application]. Once your application is approved and you have signed the agreement, we will reply with download instructions and links.
Already have Phase I access? You do not need to submit a new agreement. Simply forward your previous approval email to developer@realsee.com, including a Google-Drive-accessible email address, and we will grant you download access to the Phase II data directly.
For a detailed breakdown of dataset statistics, please refer to metadata/README.md.
- 2026-07-01: Our new work Argus — a metric panoramic 3D reconstruction method trained on Realsee3D (ECCV 2026) — is now released. The covisibility score matrix data used by Argus is now available in Realsee3D. See DATASET_STRUCTURE.md for more details.
- 2026-06-25: Phase II data released — per-viewpoint pixel-wise semantic segmentation (
segment.png) for both real-world and synthetic scenes. Real-world maps are current-model predictions (not manual annotation); ground-truth annotations are planned for a future release. See DATASET_STRUCTURE.md for details. - 2025-12-05: Phase I data(RGB-D pano and extrinsics) released.
- 2025-11-28: Dataset introduction and official website release.
Argus: Metric Panoramic 3D Reconstruction for Indoor Scenes (ECCV 2026)
Argus is a data-driven feed-forward network trained on Realsee3D that reconstructs complete, geometrically consistent, metric-scale indoor 3D scenes from sparse, unordered panoramic images in a single forward pass. Argus achieves state-of-the-art metric performance on camera pose estimation, depth estimation, and point cloud reconstruction on the Realsee3D benchmark, demonstrating the value and usability of the dataset for panoramic 3D reconstruction research.
- Project Page: https://argus-paper.realsee.ai/
- Paper: arXiv:2606.30047
- Code: github.com/realsee-developer/Argus
- Model: RealseeTechnology/argus-realsee3d
- Demo: RealseeDeveloper/Argus
Building on Realsee3D? Feel free to open a PR/issue to have your work listed here.
If you use the Realsee3D dataset in your research, please cite our paper:
@misc{Li2025realsee3d_data,
doi = {10.5281/zenodo.17826242},
url = {https://doi.org/10.5281/zenodo.17826242},
author = {Li, Linyuan and Wu, Yan and Li, Xi and Wang, Lingli and Rao, Tong and Zhou, Jie and Pan, Cihui and Hui, Xinchen},
title = {Realsee3D: A Large-Scale Multi-View RGB-D Dataset of Indoor Scenes (Version 1.1)},
publisher = {Zenodo},
year = {2026}
}
@misc{li2026argusmetricpanoramic3d,
title={Argus: Metric Panoramic 3D Reconstruction for Indoor Scenes},
author={Xi Li and Linyuan Li and Yan Wu and Tong Rao and Kai Zhang and Xinchen Hui and Cihui Pan},
year={2026},
eprint={2606.30047},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2606.30047}
}
The accompanying code for data parsing, visualization, and evaluation is released under the MIT License, allowing for free use, modification, and distribution.