Nerf2nerf
lastmod 2026-11-10created 2022-11-094 edits
nerf2nerf is a method for registering two NeRFs (Neural Radiance Fields) of the same object trained on different scenes. To estimate the relative rigid transformation between the two NeRFs, it defines a surface field derived from the density field and optimizes to align these fields. Its difference from conventional 3D registration is that it aligns the neural representations themselves without converting them to point clouds or meshes.
Pairwise Registration of Neural Radiance Fields
Documentation
- https://arxiv.org/abs/2211.01600
- [2211.01600] nerf2nerf: Pairwise Registration of Neural Radiance Fields