Incremental Multi-Scene Modelling via Continual Neural Graphics Primitives [BMVC 2025]

[Paper]    [Website]

Conditional modification to NGP

Results

Nerf Real Synthetic 360
Nerf Real Forward Facing
Tanks and Temple

Mix Scene Shared NGP Space Visualization

Image
GIF

Style Editing

Python libraries [reference]

• Create a conda environment from shared environment.yml

• Install torch-scatter following their instruction

• Install tinycudann following their instruction (pytorch extension)

• Install apex following their instruction

• Install core requirements by pip install -r requirements.txt

• Cuda extension: Upgrade pip to >= 22.1 and run pip install models/csrc/ (please run this each time you pull the code)

Note

• We will release the checkpoints later. Due to the large size, we have to upload it on the drive, and it is not possible to share it without breaking anonymity.

References

  [1] Müller, T., Evans, A., Schied, C., & Keller, A. (2022). Instant neural graphics primitives 
  with a multiresolution hash encoding. ACM transactions on graphics (TOG), 41(4), 1-15.
  [2] InstantNGP Pytorch Implementation: https://github.com/kwea123/ngp_pl]

Cite

 @article{singh2024incremental,
  title={Incremental Multi-Scene Modeling via Continual Neural Graphics Primitives},
  author={Singh, Prajwal and Tiwari, Ashish and Vashishtha, Gautam and Raman, Shanmuganathan},
  journal={arXiv preprint arXiv:2411.19903},
  year={2024}
}