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

TL;DR

The paper tackles inefficiency in Gaussian Splatting for object-specific tasks by introducing segmentation mask guided, object-centric optimization and an occlusion-aware pruning strategy. It extends 2D Gaussian Splatting with a background loss to suppress background Gaussians and enables direct mesh extraction via TSDF, yielding compact object models. The proposed method delivers object-centric Gaussian and mesh representations that are significantly smaller ($up to$ $96\%$) and faster ($up to$ $71\%$) to train while preserving quality, with immediate usability for downstream tasks such as appearance editing and physics simulation. This work enables practical, object-focused reconstructions that integrate smoothly with traditional graphics pipelines and applications.

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.

Figures (2)

• Figure 1: Our method, optimizes 2D Gaussians to accurately model specific object surfaces. They can be rendered directly or exported as a mesh. For ease of viewing, the mask is inverted and the rendered image's background is edited to be white.
• Figure :