README.md

---
pretty_name: DirectLStudio Demo Scenes
license: other
license_name: mixed-cc-by-nc-4.0-and-cc-by-4.0
license_link: https://huggingface.co/datasets/Royalvice/DirectLStudio-Demo-Scenes/blob/main/LICENSES.md
tags:
  - 3d
  - gaussian-splatting
  - light-field-display
  - directl
size_categories:
  - n<1K
---

# DirectLStudio Demo Scenes

This public dataset contains three trained 3D Gaussian Splatting scenes in the
runtime-ready SOG format used by
[DirectLStudio](https://github.com/CoronaEngine/DirectLStudio). It contains
inference assets only: no source photographs, training images, camera training
sets, checkpoints, optimizer state, or training code are included.

## Files

| Scene | Runtime file | Size | SHA-256 | Terms |
|---|---|---:|---|---|
| Lego | `scenes/lego/scene.sog` | 5,241,797 bytes | `1628DE34FA07E01960E4CD178BE7A67504C8B14A7FF46C0E396B01256EBB2756` | CC BY-NC 4.0 |
| Garden | `scenes/garden/scene.sog` | 82,435,453 bytes | `2F8DDF2D7AC131584D3AD9144C56AE6BB0D65C065D3E47EAA328636C61401910` | CC BY 4.0 |
| Bicycle | `scenes/bicycle/scene.sog` | 82,698,972 bytes | `94FEDED46A825E5DEA6C7B8D49A68DBFE9445E41947322A16E07F718F91C38AD` | CC BY 4.0 |

The machine-readable counterpart is [`manifest.json`](manifest.json). The
mixed per-file terms are authoritative in [`LICENSES.md`](LICENSES.md); the
repository-level `license: other` metadata intentionally does not flatten them
into a single license.

## Download

Install the Hugging Face CLI, then download one scene:

```powershell
hf download Royalvice/DirectLStudio-Demo-Scenes `
  --repo-type dataset `
  --include "scenes/lego/*" `
  --local-dir .\DirectLStudio-Demo-Scenes
```

DirectLStudio also provides a hash-pinned helper:

```powershell
python .\tools\download_assets.py --scene lego --output .\assets\scenes
python .\tools\download_assets.py --scene garden --output .\assets\scenes
python .\tools\download_assets.py --scene bicycle --output .\assets\scenes
```

## Provenance

- **Lego** derives from the NeRF Synthetic Lego scene published with
  [NeRF](https://www.matthewtancik.com/nerf). A trained 3DGS representation was
  converted to SOG for DirectLStudio inference. The source asset attribution
  and non-commercial restriction remain in force after training and format
  conversion.
- **Garden** and **Bicycle** derive from the
  [Mip-NeRF 360](https://jonbarron.info/mipnerf360/) dataset. Their trained 3DGS
  representations were converted to SOG for DirectLStudio inference.

These runtime files are provided for renderer evaluation and demonstration.
They do not transfer rights beyond the terms listed in `LICENSES.md`.

## Citations

```bibtex
@inproceedings{mildenhall2020nerf,
  title={NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis},
  author={Mildenhall, Ben and Srinivasan, Pratul P. and Tancik, Matthew and Barron, Jonathan T. and Ramamoorthi, Ravi and Ng, Ren},
  booktitle={ECCV},
  year={2020}
}

@article{barron2022mipnerf360,
  title={Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields},
  author={Barron, Jonathan T. and Mildenhall, Ben and Verbin, Dor and Srinivasan, Pratul P. and Hedman, Peter},
  journal={CVPR},
  year={2022}
}
```