TY - JOUR
T1 - Neural Radiance Fields in 3D Vision
T2 - A Comprehensive Review
AU - Gao, Kyle
AU - Gao, Yina
AU - He, Hongjie
AU - Lu, Dening
AU - Xu, Linlin
AU - Li, Jonathan
N1 - Publisher Copyright:
© 2024 Tsinghua University Press.
PY - 2026/8/1
Y1 - 2026/8/1
N2 - Presented in March 2020, neural radiance fields (NeRFs) have revolutionized computer vision, allowing for implicit, neural network-based scene representation, and novel view synthesis. NeRF models have found diverse applications in robotics, urban mapping, autonomous navigation, virtual and augmented reality, and more. In August 2023, Gaussian splatting, a direct competitor to NeRF-based volume rendering frameworks, was proposed. Gaussian splatting has gained tremendous momentum, overtaking NeRF-based methods as the dominant framework for novel view synthesis. We present a comprehensive survey of NeRF papers from the past five years (2020–2025). These include papers from the pre-Gaussian splatting era, where NeRFs and neural field rendering dominated the field of novel view synthesis and applications thereof. We also include works from the post-Gaussian splatting era, where NeRFs and implicit/hybrid neural fields have found niche applications. Our survey is organized into architecture and application-based taxonomies in the pre-Gaussian splatting era, as well as providing a categorization of active research areas for NeRFs, neural fields, and implicit/hybrid neural representation methods. In the post-Gaussian splatting era, we focus on relevant developments and applications. We provide an introduction to the theory of NeRFs and their training via differentiable volume rendering. We also present a benchmark comparison of the performance and speed of classical NeRFs, implicit and hybrid neural representations, and neural field models, as well as an overview of key datasets.
AB - Presented in March 2020, neural radiance fields (NeRFs) have revolutionized computer vision, allowing for implicit, neural network-based scene representation, and novel view synthesis. NeRF models have found diverse applications in robotics, urban mapping, autonomous navigation, virtual and augmented reality, and more. In August 2023, Gaussian splatting, a direct competitor to NeRF-based volume rendering frameworks, was proposed. Gaussian splatting has gained tremendous momentum, overtaking NeRF-based methods as the dominant framework for novel view synthesis. We present a comprehensive survey of NeRF papers from the past five years (2020–2025). These include papers from the pre-Gaussian splatting era, where NeRFs and neural field rendering dominated the field of novel view synthesis and applications thereof. We also include works from the post-Gaussian splatting era, where NeRFs and implicit/hybrid neural fields have found niche applications. Our survey is organized into architecture and application-based taxonomies in the pre-Gaussian splatting era, as well as providing a categorization of active research areas for NeRFs, neural fields, and implicit/hybrid neural representation methods. In the post-Gaussian splatting era, we focus on relevant developments and applications. We provide an introduction to the theory of NeRFs and their training via differentiable volume rendering. We also present a benchmark comparison of the performance and speed of classical NeRFs, implicit and hybrid neural representations, and neural field models, as well as an overview of key datasets.
KW - Gaussian splatting
KW - neural radiance fields (NeRFs)
KW - novel view synthesis
KW - volume rendering
UR - https://www.scopus.com/pages/publications/105051627545
U2 - 10.26599/CVM.2026.9450535
DO - 10.26599/CVM.2026.9450535
M3 - 文献综述
AN - SCOPUS:105051627545
SN - 2096-0433
VL - 12
SP - 851
EP - 905
JO - Computational Visual Media
JF - Computational Visual Media
IS - 4
ER -