Neural Radiance Fields and Gaussian Splatting
Optimising a 3D scene representation, a neural radiance field or a set of 3D Gaussians, from posed photos so that new views can be rendered photorealistically.
Tier Grades
Tier A: Frontier AI labs
Google publishes Zip-NeRF and CamP and Apple publishes SHARP, so frontier labs lead the method work.
Tier B: Indonesian AI labs
Per-scene optimisation needs only posed photos of one scene, which an Indonesian lab can capture and process.
Tier C: Institution units such as ITB
The 4-author Gaussian Splatting paper shows a small university team can set the state of the art here.
Tier D: Organisations such as SEMAI (PoT, PoC, PoV)
Real-time rendering at 1080p lets a delivery team turn site photos into navigable digital twins for proofs of value.
Where this comes from
Anchors
- ComputeTraining compute is not in the sources read; Zip-NeRF trains 24 times faster than mip-NeRF 360, and 3D Gaussian Splatting renders at 30 frames per second or more at 1080p.
- DataNeRF needs only a set of images with known camera poses of one scene; no large training corpus is involved for per-scene optimisation.
- Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 6 authors on the NeRF paper (arXiv 2003.08934), 4 on 3D Gaussian Splatting (arXiv 2308.04079) and 5 on Zip-NeRF (arXiv 2304.06706).
- Time horizonMonths to a year.
- Shelf-life signalsGaussian splatting gave real-time rendering that grid-based NeRFs did not, Feed-forward models such as Apple SHARP build a Gaussian scene from one image in under a second, removing per-scene optimisation
- Pipeline roleTier A and Tier C publish scene representations; Tier B captures local sites; Tier D turns them into digital twin and inspection proofs of concept.
Where this comes from
Who and What
Links
Papers
- NeRF: Representing Scenes as Neural Radiance Fields for View SynthesisBen Mildenhall et al., ECCV 2020 (per arXiv comment), 2020 founding
- 2D Gaussian Splatting for Geometrically Accurate Radiance FieldsBinbin Huang et al., arXiv preprint, 2024 state of the art
- 3D Gaussian Splatting for Real-Time Radiance Field RenderingBernhard Kerbl et al., ACM Transactions on Graphics, volume 42(4), July 2023, 2023 state of the art
- Instant Neural Graphics Primitives with a Multiresolution Hash EncodingThomas Müller et al., ACM Trans. Graph. 41, 4, Article 102 (July 2022), 15 pages, 2022 state of the art
- NeRF: Neural Radiance Field in 3D Vision: A Comprehensive Review (Updated Post-Gaussian Splatting)Kyle Gao et al., arXiv preprint, 2022 survey
Where this comes from
Sources
- arxiv.org/abs/2003.08934 https://arxiv.org/abs/2003.08934
- arxiv.org/abs/2308.04079 https://arxiv.org/abs/2308.04079
- arxiv.org/abs/2304.06706 https://arxiv.org/abs/2304.06706
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Neural Radiance Fields and Gaussian Splatting", https://aurion.semai.id/topics/neural-radiance-fields-and-gaussian-splatting/, records of October 2026.