Curvature-Aware Training for Coordinate Networks

Hemanth Saratchandran, Shin-Fang Chng, Sameera Ramasinghe, Lachlan MacDonald, Simon Lucey

ICCV 2023 pp. 13328-13338

doi:10.1109/ICCV51070.2023.01226 /iccv/2023/saratchandran2023iccv-curvatureaware/

Abstract

Coordinate networks are widely used in computer vision due to their ability to represent signals as compressed, continuous entities. However, training these networks with first-order optimizers can be slow, hindering their use in real-time applications. Recent works have opted for shallow voxel-based representations to achieve faster training, but this sacrifices memory efficiency. This work proposes a solution that leverages second-order optimization methods to significantly reduce training times for coordinate networks while maintaining their compressibility. Experiments demonstrate the effectiveness of this approach on various signal modalities, such as audio, images, videos, shape and neural radiance fields (NeRF).

PDF ICCV Semantic Scholar

Cite

Text

Saratchandran et al. "Curvature-Aware Training for Coordinate Networks." International Conference on Computer Vision, 2023. doi:10.1109/ICCV51070.2023.01226

Markdown

[Saratchandran et al. "Curvature-Aware Training for Coordinate Networks." International Conference on Computer Vision, 2023.](https://mlanthology.org/iccv/2023/saratchandran2023iccv-curvatureaware/) doi:10.1109/ICCV51070.2023.01226

BibTeX

@inproceedings{saratchandran2023iccv-curvatureaware,
  title     = {{Curvature-Aware Training for Coordinate Networks}},
  author    = {Saratchandran, Hemanth and Chng, Shin-Fang and Ramasinghe, Sameera and MacDonald, Lachlan and Lucey, Simon},
  booktitle = {International Conference on Computer Vision},
  year      = {2023},
  pages     = {13328-13338},
  doi       = {10.1109/ICCV51070.2023.01226},
  url       = {https://mlanthology.org/iccv/2023/saratchandran2023iccv-curvatureaware/}
}