RGB-D Neural Radiance Fields: Local Sampling for Faster Training
A depth-guided sampling strategy for Neural Radiance Fields that accelerates training by leveraging RGB-D sensor data to focus ray sampling along depth-informed segments, achieving faster convergence than standard NeRF.
Abstract
Learning a 3D representation of a scene has been a challenging problem for decades in computer vision. Recent advancements in implicit neural representation from images using neural radiance fields(NeRF) have shown promising results. Some of the limitations of previous NeRF based methods include longer training time, and inaccurate underlying geometry. The proposed method takes advantage of RGB-D data to reduce training time by leveraging depth sensing to improve local sampling. This paper proposes a depth-guided local sampling strategy and a smaller neural network architecture to achieve faster training time without compromising quality.
Recommended citation:
@inproceedings {10.2312:egp.20221001,
booktitle = {Eurographics 2022 - Posters},
editor = {Sauvage, Basile and Hasic-Telalovic, Jasminka},
title = {{RGB-D Neural Radiance Fields: Local Sampling for Faster Training}},
author = {Dey, Arnab and Comport, Andrew I.},
year = {2022},
publisher = {The Eurographics Association},
ISSN = {1017-4656},
ISBN = {978-3-03868-171-7},
DOI = {10.2312/egp.20221001}
}Citation
Arnab Dey, Andrew I. Comport. RGB-D Neural Radiance Fields: Local Sampling for Faster Training