Neural Radiance Field (NeRF) is a popular method for synthesizing novel views of a scene from a set of input images. While NeRF has demonstrated state-of-the-art performance in several applications, it suffers from high computational requirements. Recent works have attempted to address these issues by including explicit volumetric information, which makes the optimization process difficult when fine-graining the voxel grids. In this paper, we propose an ensemble approach that combines the strengths of two NeRF models to achieve superior results compared to state-of-the-art architectures, with a similar number of parameters. Experimental results show that our ensemble approach is a promising strategy for performance enhancement, and beats vanilla approaches under the same parameter’s cardinality constraint.

Two is Better than One: Achieving High-Quality 3D Scene Modeling with a NeRF Ensemble

Di Sario F.;Renzulli R.;Tartaglione E.;Grangetto M.
2023-01-01

Abstract

Neural Radiance Field (NeRF) is a popular method for synthesizing novel views of a scene from a set of input images. While NeRF has demonstrated state-of-the-art performance in several applications, it suffers from high computational requirements. Recent works have attempted to address these issues by including explicit volumetric information, which makes the optimization process difficult when fine-graining the voxel grids. In this paper, we propose an ensemble approach that combines the strengths of two NeRF models to achieve superior results compared to state-of-the-art architectures, with a similar number of parameters. Experimental results show that our ensemble approach is a promising strategy for performance enhancement, and beats vanilla approaches under the same parameter’s cardinality constraint.
2023
Proceedings of the 22nd International Conference on Image Analysis and Processing, ICIAP 2023
ita
2023
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Springer Science and Business Media Deutschland GmbH
14234
320
331
978-3-031-43152-4
978-3-031-43153-1
https://link.springer.com/chapter/10.1007/978-3-031-43153-1_27
3D scene modeling; Compression; Ensemble; NeRF
Di Sario F.; Renzulli R.; Tartaglione E.; Grangetto M.
File in questo prodotto:
File Dimensione Formato  
DiSarioNerf23.pdf

Accesso riservato

Tipo di file: POSTPRINT (VERSIONE FINALE DELL’AUTORE)
Dimensione 10.99 MB
Formato Adobe PDF
10.99 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
DiSarioNerf23.pdf

Accesso aperto con embargo fino al 04/09/2025

Tipo di file: POSTPRINT (VERSIONE FINALE DELL’AUTORE)
Dimensione 894.68 kB
Formato Adobe PDF
894.68 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/1951387
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact