WE-GS: An In-the-wild Efficient 3D Gaussian Representation for Unconstrained Photo Collections

TL;DR

This paper introduces WE-GS, an efficient 3D Gaussian-based framework for novel view synthesis from unconstrained photo collections, featuring a residual spherical harmonic transfer module and a spatial attention mechanism for improved rendering quality and speed.

Contribution

It extends 3D Gaussian Splatting with a residual spherical harmonic transfer and a plug-and-play spatial attention module, enabling better adaptation to lighting variations and occlusions in unconstrained photos.

Findings

Outperforms existing methods in rendering quality and speed.

Effectively adapts to lighting conditions and occlusions.

Achieves high convergence and real-time rendering.

Abstract

Novel View Synthesis (NVS) from unconstrained photo collections is challenging in computer graphics. Recently, 3D Gaussian Splatting (3DGS) has shown promise for photorealistic and real-time NVS of static scenes. Building on 3DGS, we propose an efficient point-based differentiable rendering framework for scene reconstruction from photo collections. Our key innovation is a residual-based spherical harmonic coefficients transfer module that adapts 3DGS to varying lighting conditions and photometric post-processing. This lightweight module can be pre-computed and ensures efficient gradient propagation from rendered images to 3D Gaussian attributes. Additionally, we observe that the appearance encoder and the transient mask predictor, the two most critical parts of NVS from unconstrained photo collections, can be mutually beneficial. We introduce a plug-and-play lightweight spatial attention module to simultaneously predict transient occluders and latent appearance representation for each image. After training and preprocessing, our method aligns with the standard 3DGS format and rendering pipeline, facilitating seamlessly integration into various 3DGS applications. Extensive experiments on diverse datasets show our approach outperforms existing approaches on the rendering quality of novel view and appearance synthesis with high converge and rendering speed.