libgs/ffgs-demo-examples

ffgs Demo Examples

Example inputs for the ffgs interactive demo: four real-world scenes with 10 views each, and one short video. The views are a small, downscaled selection from DL3DV-10K, taken from the undistorted DL3DV-GS-960P release.

For demonstration only. All rights stay with the original authors. If you are a rights holder and want any file removed, open a discussion on this dataset's Community tab and it will be taken down.

Scenes

Directory	Content	DL3DV scene	Views
tea-room/	Indoor: tea table, chairs, shelves	028819075f0c9b19455e99a7f9a4867a160a6749151ee72ae3e87d03c49aa3d2	10, 953×537
shop-interior/	Indoor: shop shelves, counter, aisle	00713c8c22cf3b2ef6495b1da5484da9972921442d85a0a3c8be57f7aa9bbbb5	10, 956×535
courtyard/	Outdoor: stone archway, eaves, lantern	001dccbc1f78146a9f03861026613d8e73f39f372b545b26118e37a23c740d5f	10, 955×537
woodland-path/	Outdoor: stone path, trees, sculpture	008c201a7eff27ce0413f7931a48e92cf05ded9d9b7cf16cc2276ff3b80c7b22	10, 953×537
videos/courtyard.mp4	Courtyard, 55 consecutive frames	same as courtyard/	956×538, 12 fps, H.264

Scene names are descriptive labels, not official location names. manifest.json lists the source revision, source frame and SHA-256 of every file.

Processing

• Views are frames 0, 6, …, 54 of each scene's trajectory, saved as JPEG at the source resolution (no resizing or cropping).
• The video encodes frames 0–54 of the courtyard scene at 12 fps without audio. Its right and bottom edges are padded by 1 px to even dimensions.
• No camera poses are included; the files are meant as unposed image / video input.

License and attribution

The source data is licensed under CC BY-NC 4.0 together with the DL3DV-10K Terms of Use, and the same terms apply to these files. Images by the DL3DV-10K Dataset Collection Team; undistortion by the FCGS authors. If you use these files, please cite:

@inproceedings{ling2024dl3dv,
  title={DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision},
  author={Ling, Lu and Sheng, Yichen and Tu, Zhi and Zhao, Wentian and Xin, Cheng and Wan, Kun and Yu, Lantao and Guo, Qianyu and Yu, Zixun and Lu, Yawen and others},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={22160--22169},
  year={2024}
}
@inproceedings{fcgs2025,
  title={Fast Feedforward 3D Gaussian Splatting Compression},
  author={Chen, Yihang and Wu, Qianyi and Li, Mengyao and Lin, Weiyao and Harandi, Mehrtash and Cai, Jianfei},
  booktitle={The Thirteenth International Conference on Learning Representations},
  year={2025}
}