VISAPP 2026

ROI-NeRFs: Hi-Fi Visualization of Objects of Interest
within a Scene by NeRFs Composition

Quoc-Anh Bui^1,2, Gilles Rougeron¹, Géraldine Morin², Simone Gasparini²

¹Université Paris-Saclay, CEA-List ²Université de Toulouse, Toulouse INP, IRIT

arXiv Paper Code

ROI-NeRFs teaser showing enhanced detail on objects of interest — **ROI-NeRFs** decomposes a scene into a global Scene NeRF and multiple local ROI NeRFs, enabling high-fidelity visualization of selected objects of interest within large-scale cultural heritage scenes.

Abstract

Method Overview

ROI Decomposition

Camera Selection

Ray-level Composition

ROI-NeRFs Pipeline — **Figure 1.** Overview of the ROI-NeRFs pipeline.

BibTeX

@inproceedings{bui2024roinerfs,
  title     = {ROI-NeRFs: Hi-Fi Visualization of Objects of Interest within a Scene by NeRFs Composition},
  author    = {Bui, Quoc-Anh and Rougeron, Gilles and Morin, Géraldine and Gasparini, Simone},
  booktitle = {International Conference on Computer Vision Theory and Applications (VISAPP)},
  year      = {2026}
}