Abstract

The problem of inferring object shape from a single 2D image is underconstrained. Prior knowledge about what objects are plausible can help, but even given such prior knowledge there may still be uncertainty about the shapes of occluded parts of objects. Recently, conditional neural radiance field (NeRF) models have been developed that can learn to infer good point estimates of 3D models from single 2D images. The problem of inferring uncertainty estimates for these models has received less attention. In this work, we propose probabilistic NeRF (ProbNeRF), a model and inference strategy for learning probabilistic generative models of 3D objects' shapes and appearances, and for doing posterior inference to recover those properties from 2D images. ProbNeRF is trained as a variational autoencoder, but at test time we use Hamiltonian Monte Carlo (HMC) for inference. Given one or a few 2D images of an object (which may be partially occluded), ProbNeRF is able not only to accurately model the parts it sees, but also to propose realistic and diverse hypotheses about the parts it does not see. We show that key to the success of ProbNeRF are (i) a deterministic rendering scheme, (ii) an annealed-HMC strategy, (iii) a hypernetwork-based decoder architecture, and (iv) doing inference over a full set of NeRF weights, rather than just a low-dimensional code.

HMC and VI posterior sample variation

Depicted below are some HMC and VI posterior samples from the GHUM and SRN cars datasets:

Conditioned on either the left half a view of a GHUM body (top left) or a back view of an SRN car (1st column, 4th row), HMC produces samples (columns 2–4) that are realistic, consistent with the conditioned-on view, and diverse as shown by the per-pixel variance (column 5). VI produces realistic and consistent samples (columns 6–8), but they have almost no diversity (last column). Renderings from novel views (rows 2–3/5–6) highlight the diversity of the HMC samples, in e.g. the poses of the left arm and the left leg or the variation of the car’s shape, spoiler and color.

Posterior HMC samples

Prior samples

Architecture

Citation

@article{hoffman2023probnerf,
    title={ProbNeRF: Uncertainty-Aware Inference of 3D Shapes from 2D Images},
    author={Hoffman, Matthew D. and Le, Tuan Anh and Sountsov, Pavel and Suter, Christopher and Lee, Ben and Mansinghka, Vikash K. and Saurous, Rif A.},
    journal={AISTATS},
    year={2023}
}

Acknowledgements

The website template was borrowed from Mip-NeRF360.