# Instant-NeRF: Instant On-Device Neural Radiance Field Training via Algorithm-Accelerator Co-Designed Near-Memory Processing

arXiv: https://arxiv.org/abs/2305.05766 · HTML: https://mlsanity.com/paper/2305.05766/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 21 references that are in the map.
- First posted on arXiv 9 May 2023.

## Builds on (most cited first, 4 of 4)

- [Instant Neural Graphics Primitives with a Multiresolution Hash Encoding](https://mlsanity.com/paper/2201.05989/) (2022)
- [TensoRF: Tensorial Radiance Fields](https://mlsanity.com/paper/2203.09517/) (2022)
- [FastNeRF: High-Fidelity Neural Rendering at 200FPS](https://mlsanity.com/paper/2103.10380/) (2021)
- [TensorDIMM: A Practical Near-Memory Processing Architecture for Embeddings and Tensor Operations in Deep Learning](https://mlsanity.com/paper/1908.03072/) (2019)

## Built on by (most cited first, 6 of 6)

- [GS-2DGS: Geometrically Supervised 2DGS for Reflective Object Reconstruction](https://mlsanity.com/paper/2506.13110/) (2025)
- [SpNeRF: Memory Efficient Sparse Volumetric Neural Rendering Accelerator for Edge Devices](https://mlsanity.com/paper/2505.08191/) (2025)
- [DreamSat-2.0: Towards a General Single-View Asteroid 3D Reconstruction](https://mlsanity.com/paper/2508.01079/) (2025)
- [Distilled-3DGS:Distilled 3D Gaussian Splatting](https://mlsanity.com/paper/2508.14037/) (2025)
- [FACT-GS: Frequency-Aligned Complexity-Aware Texture Reparameterization for 2D Gaussian Splatting](https://mlsanity.com/paper/2511.23292/) (2025)
- [Node-RF: Learning Generalized Continuous Space-Time Scene Dynamics with Neural ODE-based NeRFs](https://mlsanity.com/paper/2603.12078/) (2026)

## Abstract (arXiv)

Instant on-device Neural Radiance Fields (NeRFs) are in growing demand for unleashing the promise of immersive AR/VR experiences, but are still limited by their prohibitive training time. Our profiling analysis reveals a memory-bound inefficiency in NeRF training. To tackle this inefficiency, near-memory processing (NMP) promises to be an effective solution, but also faces challenges due to the unique workloads of NeRFs, including the random hash table lookup, random point processing sequence, and heterogeneous bottleneck steps. Therefore, we propose the first NMP framework, Instant-NeRF, dedicated to enabling instant on-device NeRF training. Experiments on eight datasets consistently validate the effectiveness of Instant-NeRF.