LUDVIG: Learning-Free Uplifting of 2D Visual Features to Gaussian Splatting Scenes
Juliette Marrie¹², Romain Menegaux¹, Michael Arbel¹, Diane Larlus², Julien Mairal¹
¹Inria ²NAVER LABS Europe

Contents:

- 00011_supp.pdf
  Supplementary document with additional implementation details, quantitative and qualitative results.

- 00011_diffusion.mp4
  Visualization of the graph diffusion process used to perform 3D segmentation from scribbles based on DINOv2 feature similarity (Fern scene, NVOS dataset).

- 00011_figurines.mp4, 00011_teatime.mp4
  Open-vocabulary 3D segmentation from 3D CLIP similarities refined with DINOv2-based graph diffusion (Figurine and Teatime scenes, LERF dataset).

BibTeX:

@inproceedings{marrie2025ludvig,
  title={LUDVIG: Learning-Free Uplifting of 2D Visual features to Gaussian Splatting Scenes},
  author={Marrie, Juliette and Menegaux, Romain and Arbel, Michael and Larlus, Diane and Mairal, Julien},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2025}
}
