NeRF and radiance fields: research map
733 papers on NeRF and radiance fields in 3D vision, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 7 approaches. The busiest year so far is 2023.
What it is
A neural radiance field (NeRF) represents a scene as a neural network that gives colour and density at every point in space; rendering it from a new viewpoint produces realistic images. Research makes NeRFs work from few photos, for moving scenes and for editing, and recovers materials and lighting.
Terms you will see on this page:
- NeRF
- neural radiance field.
- Volume rendering
- combining colour and density along each camera ray to form a pixel.
- Inverse rendering
- recovering geometry, materials and lighting from images.
Start reading with the most cited papers here:
- Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022 · 1,901 citations)
- pixelNeRF: Neural Radiance Fields From One or Few Images (CVPR 2021 · 1,530 citations)
Within 3D vision, its share shrank from 14.9% in 2023–24 to 3.3% in 2025–26 (266 → 90 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore NeRF and radiance fields in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Inverse rendering with neural fields rising: 18% → 37% of the topic (48 → 33 papers)
- NeRF for scene decomposition and understanding rising: 6% → 11% of the topic (17 → 10 papers)
- Dynamic and editable NeRF falling: 23% → 14% of the topic (60 → 13 papers)
- Single-view and streaming radiance fields falling: 12% → 6% of the topic (31 → 5 papers)
Approaches and key papers
NeRF from few views, dynamic and editable scenes · 484 papers
Keywords: radiance fields, fields nerf, nerfs
Approaches in this cluster:
- Sparse-input NeRF regularization (166 papers)
Improves NeRF view synthesis with geometry, depth and ray priors, and in-the-wild robustness. - Dynamic and editable NeRF (165 papers)
Extends radiance fields to dynamic scenes, editing, large scenes and faster real-time rendering. - Single-view and streaming radiance fields (95 papers)
Builds conditioned, recolorable or streaming radiance fields from limited or sequential observations. - NeRF for scene decomposition and understanding (58 papers)
Uses radiance fields for geometry decomposition, object selection, segmentation and detection.
Most cited and most cited since 2024:
- Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields (CVPR 2022 · 1,901 citations)
- pixelNeRF: Neural Radiance Fields From One or Few Images (CVPR 2021 · 1,530 citations)
- ReconFusion: 3D Reconstruction with Diffusion Priors (CVPR 2024 · 185 citations)
- Bayes' Rays: Uncertainty Quantification for Neural Radiance Fields (CVPR 2024 · 58 citations)
Neural rendering and inverse rendering · 249 papers
Keywords: rendering, field, synthesis
Approaches in this cluster:
- Inverse rendering with neural fields (152 papers)
Recovers geometry, reflectance and lighting from multi-view images for relighting and view synthesis. - Few-shot neural rendering architectures (61 papers)
Designs light-field, attention-based and reposable representations for view synthesis from few views. - 3D-aware generative radiance fields (36 papers)
Trains GAN-based radiance fields to generate 3D-consistent images.
Most cited and most cited since 2024:
- GIRAFFE: Representing Scenes As Compositional Generative Neural Feature Fields (CVPR 2021 · 742 citations)
- IBRNet: Learning Multi-View Image-Based Rendering (CVPR 2021 · 722 citations)
- DL3DV-10K: A Large-Scale Scene Dataset for Deep Learning-based 3D Vision (CVPR 2024 · 140 citations)
- Structure-Aware Sparse-View X-ray 3D Reconstruction (CVPR 2024 · 77 citations)
Related topics in 3D vision
- Gaussian splatting (1,340)
- Human pose estimation (2,063)
- 3D generation (2,448)
- Face and shape (2,246)
- Depth and stereo (2,574)
