Extending DeepSDF for automatic 3D shape retrieval and similarity transform estimation

TL;DR

This paper extends DeepSDF to enable automatic 3D shape retrieval and similarity transform estimation, allowing for effective shape comparison and compression in real-world scenarios.

Contribution

It introduces a joint estimation method for shape and transform parameters, improving DeepSDF's applicability to real-world 3D shape retrieval.

Findings

Effective on synthetic and real datasets

Outperforms state-of-the-art methods

Supports 3D model compression

Abstract

Recent advances in computer graphics and computer vision have found successful application of deep neural network models for 3D shapes based on signed distance functions (SDFs) that are useful for shape representation, retrieval, and completion. However, this approach has been limited by the need to have query shapes in the same canonical scale and pose as those observed during training, restricting its effectiveness on real world scenes. We present a formulation to overcome this issue by jointly estimating shape and similarity transform parameters. We conduct experiments to demonstrate the effectiveness of this formulation on synthetic and real datasets and report favorable comparisons to the state of the art. Finally, we also emphasize the viability of this approach as a form of 3D model compression.