kkennethwu/360-USID

Introduction

We present the 360-USID dataset in AuraFusion360: Augmented Unseen Region Alignment for Reference-based 360° Unbounded Scene Inpainting.

We introduce first 360° Unbounded Scenes Inpaint- ing Dataset (360-USID), consisting of seven scenes with training views (RGB images and object masks), novel test- ing views (inpainting ground truth), camera poses, and a reference view (without objects) for evaluating with other reference-based methods.

We further collect and process some 360 unbounded scenes in Other-360 for qualitative comparison. These scenes are from Mip-NeRF 360, NeRF-Real-360, Instruct-NeRF2NeRF.

Download

huggingface-cli login
huggingface-cli download kkennethwu/360-USID --repo-type dataset --local-dir ./data --resume-download --quiet --max-workers 32

Data tree

360-USID:

360-USID/{scene}
    ├── images
    ├── object_masks
    ├── reference
    ├── sparse
    ├── test_images # gt images with objects removed.
    ├── test_object_masks # object masks in testing views rendered by Masked-GS, used for evaluation.
    └── unseen_masks

Other-360:

Other-360/{scene}
    ├── images
    ├── object_masks
    ├── reference
    ├── sparse
    └── unseen_masks

Citation

If you find this dataset helpful, please cite this paper and give us a ⭐️.

@misc{wu2025aurafusion360augmentedunseenregion,
      title={AuraFusion360: Augmented Unseen Region Alignment for Reference-based 360{\deg} Unbounded Scene Inpainting}, 
      author={Chung-Ho Wu and Yang-Jung Chen and Ying-Huan Chen and Jie-Ying Lee and Bo-Hsu Ke and Chun-Wei Tuan Mu and Yi-Chuan Huang and Chin-Yang Lin and Min-Hung Chen and Yen-Yu Lin and Yu-Lun Liu},
      year={2025},
      eprint={2502.05176},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2502.05176}, 
}