DeepAI AI Chat
Log In Sign Up

GenSDF: Two-Stage Learning of Generalizable Signed Distance Functions

by   Gene Chou, et al.
Princeton University

We investigate the generalization capabilities of neural signed distance functions (SDFs) for learning 3D object representations for unseen and unlabeled point clouds. Existing methods can fit SDFs to a handful of object classes and boast fine detail or fast inference speeds, but do not generalize well to unseen shapes. We introduce a two-stage semi-supervised meta-learning approach that transfers shape priors from labeled to unlabeled data to reconstruct unseen object categories. The first stage uses an episodic training scheme to simulate training on unlabeled data and meta-learns initial shape priors. The second stage then introduces unlabeled data with disjoint classes in a semi-supervised scheme to diversify these priors and achieve generalization. We assess our method on both synthetic data and real collected point clouds. Experimental results and analysis validate that our approach outperforms existing neural SDF methods and is capable of robust zero-shot inference on 100+ unseen classes. Code can be found at


page 7

page 18

page 19

page 20


Unsupervised Inference of Signed Distance Functions from Single Sparse Point Clouds without Learning Priors

It is vital to infer signed distance functions (SDFs) from 3D point clou...

OSSGAN: Open-Set Semi-Supervised Image Generation

We introduce a challenging training scheme of conditional GANs, called o...

Learning to Learn in a Semi-Supervised Fashion

To address semi-supervised learning from both labeled and unlabeled data...

Theoretical and empirical analysis of a fast algorithm for extracting polygons from signed distance bounds

We investigate an asymptotically fast method for transforming signed dis...

MetaSets: Meta-Learning on Point Sets for Generalizable Representations

Deep learning techniques for point clouds have achieved strong performan...

Meta Discovery: Learning to Discover Novel Classes given Very Limited Data

In learning to discover novel classes(L2DNC), we are given labeled data ...

Spacing Loss for Discovering Novel Categories

Novel Class Discovery (NCD) is a learning paradigm, where a machine lear...