| --- |
| pretty_name: msg |
| task_categories: |
| - graph-ml |
| --- |
| |
| # Multiview Scene Graph (NeurIPS 2024) |
| This is the dataset for: |
| > [**Multiview Scene Graph**](https://www.ai4ce.github.io/MSG) |
| > |
| > Juexiao Zhang, Gao Zhu, Sihang Li, Xinhao Liu, Haorui Song, Xinran Tang, Chen Feng |
| > |
| > New York University |
|
|
| [[arXiv]](https://arxiv.org/abs/2410.11187) |
|
|
| # HSG: Hyperbolic Scene Graph |
| This dataset is also used in the paper: |
| > [**HSG: Hyperbolic Scene Graph**](https://huggingface.co/papers/2604.17454) |
| > |
| > Liyang Wang, Zeyu Zhang, and Hao Tang |
|
|
| [[Code]](https://github.com/AIGeeksGroup/HSG) |
|
|
| ## Introduction |
| HSG is a **hyperbolic representation learning framework** that models scene graphs in non-Euclidean space to better capture hierarchical relationships for structured 3D scene understanding. |
|
|
| HSG data is converted from Apple's [ARKitScenes](https://github.com/apple/ARKitScenes) by transforming its 3D annotations to 2D. |
|
|
| ## License |
| This dataset is based on Apple's [ARKitScenes](https://github.com/apple/ARKitScenes/tree/main) dataset, so please obey their [license](https://github.com/apple/ARKitScenes/blob/main/LICENSE). |
|
|
| ## Citation |
| If you find this code or paper helpful, please consider citing: |
| ```bibtex |
| @article{wang2026hsg, |
| title={HSG: Hyperbolic Scene Graph}, |
| author={Wang, Liyang and Zhang, Zeyu and Tang, Hao}, |
| year={2026} |
| } |
| |
| @inproceedings{zhang2024multiview, |
| title={Multiview Scene Graph}, |
| author={Zhang, Juexiao and Zhu, Gao and Li, Sihang and Liu, Xinhao fabric and Song, Haorui and Tang, Xinran and Feng, Chen}, |
| booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems}, |
| year={2024} |
| } |
| ``` |
