Direct Object-Level Reconstruction via Probabilistic Gaussian Splatting

TL;DR

This paper introduces a novel single-object 3D reconstruction method using probabilistic Gaussian splatting, which improves efficiency by focusing on the object of interest and dynamically pruning background information, while maintaining high reconstruction quality.

Contribution

The method integrates foreground-background probability cues into Gaussian primitives and employs a dual-stage filtering strategy, enabling efficient and robust single-object 3D reconstruction with reduced Gaussian count.

Findings

Achieves comparable quality to standard 3D Gaussian Splatting with only 1/10 of the Gaussians.

Demonstrates strong self-correction capability with mask errors.

Requires significantly less memory and computation than full-scene approaches.

Abstract

Object-level 3D reconstruction play important roles across domains such as cultural heritage digitization, industrial manufacturing, and virtual reality. However, existing Gaussian Splatting-based approaches generally rely on full-scene reconstruction, in which substantial redundant background information is introduced, leading to increased computational and storage overhead. To address this limitation, we propose an efficient single-object 3D reconstruction method based on 2D Gaussian Splatting. By directly integrating foreground-background probability cues into Gaussian primitives and dynamically pruning low-probability Gaussians during training, the proposed method fundamentally focuses on an object of interest and improves the memory and computational efficiency. Our pipeline leverages probability masks generated by YOLO and SAM to supervise probabilistic Gaussian attributes, replacing binary masks with continuous probability values to mitigate boundary ambiguity. Additionally, we propose a dual-stage filtering strategy for training's startup to suppress background Gaussians. And, during training, rendered probability masks are conversely employed to refine supervision and enhance boundary consistency across views. Experiments conducted on the MIP-360, T&T, and NVOS datasets demonstrate that our method exhibits strong self-correction capability in the presence of mask errors and achieves reconstruction quality comparable to standard 3DGS approaches, while requiring only approximately 1/10 of their Gaussian amount. These results validate the efficiency and robustness of our method for single-object reconstruction and highlight its potential for applications requiring both high fidelity and computational efficiency.