Object-Centric 2D Gaussian Splatting: Background Removal and Occlusion-Aware Pruning for Compact Object Models

TL;DR

This paper introduces an object-centric Gaussian Splatting method that uses object masks and occlusion-aware pruning to produce compact, efficient, and high-quality 2D Gaussian models suitable for various downstream applications.

Contribution

It presents a novel object-centric reconstruction approach with occlusion-aware pruning, significantly reducing model size and training time while maintaining quality.

Findings

Models are up to 96% smaller.

Training is up to 71% faster.

Retains competitive quality.

Abstract

Current Gaussian Splatting approaches are effective for reconstructing entire scenes but lack the option to target specific objects, making them computationally expensive and unsuitable for object-specific applications. We propose a novel approach that leverages object masks to enable targeted reconstruction, resulting in object-centric models. Additionally, we introduce an occlusion-aware pruning strategy to minimize the number of Gaussians without compromising quality. Our method reconstructs compact object models, yielding object-centric Gaussian and mesh representations that are up to 96% smaller and up to 71% faster to train compared to the baseline while retaining competitive quality. These representations are immediately usable for downstream applications such as appearance editing and physics simulation without additional processing.