Instant Uncertainty Calibration of NeRFs Using a Meta-calibrator

arXiv:2312.02350v2 Announce Type: replace
Abstract: Although Neural Radiance Fields (NeRFs) have markedly improved novel view synthesis, accurate uncertainty quantification in their image predictions remains an open problem. The prevailing methods for estimating uncertainty, including the state-of-the-art Density-aware NeRF Ensembles (DANE) [29], quantify uncertainty without calibration. This frequently leads to over- or under-confidence in image predictions, which can undermine their real-world applications. In this paper, we propose a method which, for the first time, achieves calibrated uncertainties for NeRFs. To accomplish …

abstract art arxiv confidence cs.cv fields image instant leads meta nerf neural radiance fields novel predictions quantification state synthesis type uncertainty view