google/nerfies
A neural radiance field implementation for reconstructing 3D scenes and synthesizing novel views from multi-view image datasets.
Not currently ranked — collecting fresh signals.
star history
Deformable Neural Radiance Fields (Nerfies) reconstruct 3D scenes from collections of casual photos by learning neural implicit representations. The method extends NeRF to handle non-rigid deformations, enabling reconstruction of deformable objects like captured selfie videos. Built on JAX and JaxNeRF, the codebase supports training custom Nerfie models using GPUs or Cloud TPUs, with Colab notebooks provided for easy experimentation.
Frequently asked
- What is google/nerfies?
- A neural radiance field implementation for reconstructing 3D scenes and synthesizing novel views from multi-view image datasets.
- Is nerfies open source?
- Yes — google/nerfies is open source, released under the Apache-2.0 license.
- What language is nerfies written in?
- google/nerfies is primarily written in Python.
- How popular is nerfies?
- google/nerfies has 2k stars on GitHub.
- Where can I find nerfies?
- google/nerfies is on GitHub at https://github.com/google/nerfies.