Key points
- A multiresolution hash encoding reduces NeRF training from hours to seconds on a single GPU by replacing coordinate-based networks with a compact hash table lookup.
- It won the SIGGRAPH 2022 Best Paper award and was named a Time Best Invention of 2022.
- Instant-NGP's speed gains made neural scene capture production-feasible and directly enabled Nerfstudio and 3D Gaussian splatting as subsequent production-pipeline tools.
Summary
Mueller et al. (2022) introduce a multiresolution hash encoding that replaces the large fully connected networks used in earlier NeRF methods, reducing training time from hours to seconds on a single consumer GPU. The result made neural scene capture practically accessible and is the direct technical antecedent of Nerfstudio and the subsequent 3D Gaussian splatting literature. For animation and VFX education, the paper marks the moment neural capture became teachable as a production workflow rather than a research experiment.
Related items
- Nerfstudio: A Modular Framework for Neural Radiance Field Development
- 3D Gaussian Splatting for Real-Time Radiance Field Rendering
Source
Source: ACM Transactions on Graphics (SIGGRAPH 2022) ↗ (Research)
Cite this item
Mueller, T., Evans, A., Schied, C. & Keller, A. (2022). ‘Instant Neural Graphics Primitives with a Multiresolution Hash Encoding’, ACM Transactions on Graphics (SIGGRAPH 2022). Available at: https://nvlabs.github.io/instant-ngp/
Your reference manager can also read this page directly: with the Zotero (or Mendeley) browser connector installed, save it straight to your library. Whole-collection exports: RIS, BibTeX, CSL-JSON.