MetaSDF: Meta-learning Signed Distance Functions

By Vincent Sitzmann, Eric Chan, Richard Tucker, Noah Snavely, Gordon Wetzstein. Neural Information Processing Systems, 2020

295 citations recorded by the source corpus.

Neural implicit shape representations are an emerging paradigm that offers many potential benefits over conventional discrete representations, including memory efficiency at a high spatial resolution. Generalizing across shapes with such neural implicit representations amounts to learning priors over the respective function space and enables geometry reconstruction from partial or noisy observations. Existing generalization methods rely on conditioning a neural network on a low-dimensional latent code that is either regressed by an encoder or jointly optimized in the auto-decoder framework. Here, we formalize learning of a shape space as a meta-learning problem and leverage gradient-based meta-learning algorithms to solve this task. We demonstrate that this approach performs on par with auto-decoder based approaches while being an order of magnitude faster at test-time inference. We further demonstrate that the proposed gradient-based method outperforms encoder-decoder based methods that leverage pooling-based set encoders.

Subfields: Computational Mechanics, Computer Graphics and Computer-Aided Design

View Paper: https://www.semanticscholar.org/paper/ebae0a96098bb84eee418c76f5b6a2cb87614ac1

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Graphics and Computer-Aided Design

Citations in the corpus

Putting NeRF on a Diet: Semantically Consistent Few-Shot View Synthesis (2021, background): “Similarly, [34] meta-learns a signed distance field for shape representation problems.”
Generative Novel View Synthesis with 3D-Aware Diffusion Models (2023, methodology): “[67] Vincent Sitzmann, Eric Chan, Richard Tucker, Noah Snavely, and Gordon Wetzstein.”
Advances in Neural Rendering (2021, background): “An alternative to representing the scene as a low-dimensional latent code is to quickly optimize the weights of an MLP scene representation in a few optimization steps via gradient-based meta-learning [SCT ∗ 20].”
Light Field Networks: Neural Scene Representations with Single-Evaluation Rendering (2021, methodology): “We do not claim that this particular meta-learning method will out-perform other forms of conditioning, such as gradient-based meta-learning [57, 23] or FILM conditioning [58], but perform a comparison to a conditioning-by-concatenation approach in the appendix.”
Learning Continuous Image Representation with Local Implicit Image Function (2020, background): “[40] also proposes a meta-learning-based method for sharing the function space.”
Neural Descriptor Fields: SE(3)-Equivariant Object Representations for Manipulation (2021, background): “Their functional nature further enables the construction of latent spaces that encode class information as well as 3D correspondence [8, 17, 39].”