Single-View 3D Reconstruction via SO(2)-Equivariant Gaussian Sculpting Networks

TL;DR

This paper presents SO(2)-Equivariant Gaussian Sculpting Networks (GSNs) for efficient single-view 3D object reconstruction, achieving high throughput and competitive quality, with applications in robotic manipulation.

Contribution

Introduction of GSNs that leverage SO(2)-equivariance for fast, high-quality 3D reconstruction from a single view, integrating Gaussian representations and multi-view training.

Findings

Achieves >150FPS reconstruction speed.

Performs comparably to diffusion-based methods in quality.

Effective in robotic object grasping tasks.

Abstract

This paper introduces SO(2)-Equivariant Gaussian Sculpting Networks (GSNs) as an approach for SO(2)-Equivariant 3D object reconstruction from single-view image observations. GSNs take a single observation as input to generate a Gaussian splat representation describing the observed object's geometry and texture. By using a shared feature extractor before decoding Gaussian colors, covariances, positions, and opacities, GSNs achieve extremely high throughput (>150FPS). Experiments demonstrate that GSNs can be trained efficiently using a multi-view rendering loss and are competitive, in quality, with expensive diffusion-based reconstruction algorithms. The GSN model is validated on multiple benchmark experiments. Moreover, we demonstrate the potential for GSNs to be used within a robotic manipulation pipeline for object-centric grasping.