Abstract:Reconstructing large-scale scenes using Neural Radiance Fields (NeRFs) is a research hotspot in 3D computer vision. Existing MLP (multi-layer perception)-based methods often suffer from issues of underfitting and a lack of fine details in rendering large-scale scenes. Popular solutions are to divide the scene into small areas for separate modeling or to increase the layer scale of the MLP network. However, the subsequent problem is that the training cost increases. Moreover, reconstructing large scenes, unlike object-scale reconstruction, involves a geometrically considerable increase in the quantity of view data if the prior information of the scene is not effectively utilized. In this paper, we propose an innovative method named MM-NeRF, which integrates efficient hybrid features into the NeRF framework to enhance the reconstruction of large-scale scenes. We propose employing a dual-branch feature capture structure, comprising a multi-resolution 3D hash grid feature branch and a multi-view 2D prior feature branch. The 3D hash grid feature models geometric details, while the 2D prior feature supplements local texture information. Our experimental results show that such integration is sufficient to render realistic novel views with fine details, forming a more accurate geometric representation. Compared with representative methods in the field, our method significantly improves the PSNR (Peak Signal-to-Noise Ratio) by approximately 5%. This remarkable progress underscores the outstanding contribution of our method in the field of large-scene radiance field reconstruction.

DistGrid: Scalable Scene Reconstruction with Distributed Multi-resolution Hash Grid

MM-NeRF: Large-Scale Scene Representation with Multi-Resolution Hash Grid and Multi-View Priors Features

Scalable Neural Indoor Scene Rendering.

Grid-guided Neural Radiance Fields for Large Urban Scenes

Efficient large-scale scene representation with a hybrid of high-resolution grid and plane features

MD-NeRF: Enhancing Large-Scale Scene Rendering and Synthesis With Hybrid Point Sampling and Adaptive Scene Decomposition

ScaNeRF: Scalable Bundle-Adjusting Neural Radiance Fields for Large-Scale Scene Rendering

HollowNeRF: Pruning Hashgrid-Based NeRFs with Trainable Collision Mitigation

NeRFusion: Fusing Radiance Fields for Large-Scale Scene Reconstruction

Dynamic Neural Radiosity with Multi-grid Decomposition

VastGaussian: Vast 3D Gaussians for Large Scene Reconstruction

Digging into radiance grid for real-time view synthesis with detail preservation

SCALAR-NeRF: SCAlable LARge-scale Neural Radiance Fields for Scene Reconstruction

Neural Radiance Fields with Hash-Low-Rank Decomposition

Adaptive Multi-NeRF: Exploit Efficient Parallelism in Adaptive Multiple Scale Neural Radiance Field Rendering

Drone-NeRF: Efficient NeRF based 3D scene reconstruction for large-scale drone survey

PyNeRF: Pyramidal Neural Radiance Fields

GaRField++: Reinforced Gaussian Radiance Fields for Large-Scale 3D Scene Reconstruction