WildCity

Yuto Furutani*, Takashi Otonari*, Kaede Shiohara*, Toshihiko Yamasaki
The University of Tokyo. *Co-first authors.

Overview

Our WildCity dataset is designed to train feed-forward 3DGS models for 3D reconstruction from unconstrained photo collections. The dataset comprises multi-view, multi-lighting images covering 200 scenes, 170 HDRI maps, and diverse transient objects.

Dataset Structure

license.csv                        # Attribution info for all 3D models used
scene_city_mapping.csv             # Mapping from scene names to city IDs
wildcity/                          # Main dataset: 200 scenes
├── 0_visualization/               # Overview PDFs for all scenes
└── Scene{1..200}/
    ├── Cam_{000..049}/            # 50 camera viewpoints per scene
    │   ├── light_{000..029}.png   # 30 illumination conditions
    │   ├── depth.exr              # Depth map
    │   └── mask.png               # Sky mask
    └── colmap/
        ├── cameras.txt            # Camera intrinsics
        └── images.txt             # Camera extrinsics
wildcity_aug/                      # Augmented subset
└── Scene{...}/
    └── Cam_{000..049}/
        └── light_{000..029}.png   # Randomly sampled viewpoints

Image Resolutions

• wildcity: 512x512 pixels
• wildcity_aug: 1024x1024 pixels

wildcity contains the original multi-illumination renders.
wildcity_aug is a randomly sampled subset (1/8 of wildcity scenes) with transient objects inserted via Gemini. Where the same scene, camera, and illumination exist in both splits (e.g., wildcity/Scene1/Cam_000/light_003.png and wildcity_aug/Scene1/Cam_000/light_003.png), the images differ only in the presence of transient objects.

License

Licenses vary by city, as the underlying 3D assets carry different terms. Per-model details are in license.csv; scene-to-city mappings are in scene_city_mapping.csv.

City	Scenes	License
city1	Scene1–25	CC BY
city2	Scene26–50	CC BY-NC
city3	Scene51–72	CC BY-NC
city4	Scene73–96	CC BY
city5	Scene97–117	CC BY-NC
city6	Scene118–135	CC BY-SA
city7	Scene136–150	CC BY-SA
city8	Scene151–175	CC BY
city9	Scene176–200	CC BY

Citing

@article{furutani2026wild3r,
  title   = {Wild3R: Feed-Forward 3D Gaussian Splatting from Unconstrained Sparse Photo Collection},
  author  = {Furutani, Yuto and Otonari, Takashi and Shiohara, Kaede and Yamasaki, Toshihiko},
  journal = {arXiv preprint arXiv:2606.11894},
  year    = {2026}
}

Acknowledgements

We thank all authors behind these projects and repositories for their excellent work: LightCity, SceneCity, Sketchfab, and Gemini.