FaithFusion: Harmonizing Reconstruction and Generation via Pixel-wise Information Gain

TL;DR

FaithFusion introduces a pixel-wise information gain approach to effectively fuse 3D scene reconstruction with diffusion models, improving geometric fidelity and visual quality under viewpoint shifts.

Contribution

It proposes a novel pixel-wise Expected Information Gain framework for 3D scene reconstruction and generation fusion, addressing geometric drift and over-restoration issues.

Findings

Achieves state-of-the-art performance on Waymo dataset metrics.

Maintains high visual fidelity with stable FID at large viewpoint shifts.

Effective fusion without additional priors or structural changes.

Abstract

In controllable driving-scene reconstruction and 3D scene generation, maintaining geometric fidelity while synthesizing visually plausible appearance under large viewpoint shifts is crucial. However, effective fusion of geometry-based 3DGS and appearance-driven diffusion models faces inherent challenges, as the absence of pixel-wise, 3D-consistent editing criteria often leads to over-restoration and geometric drift. To address these issues, we introduce \textbf{FaithFusion}, a 3DGS-diffusion fusion framework driven by pixel-wise Expected Information Gain (EIG). EIG acts as a unified policy for coherent spatio-temporal synthesis: it guides diffusion as a spatial prior to refine high-uncertainty regions, while its pixel-level weighting distills the edits back into 3DGS. The resulting plug-and-play system is free from extra prior conditions and structural modifications.Extensive experiments on the Waymo dataset demonstrate that our approach attains SOTA performance across NTA-IoU, NTL-IoU, and FID, maintaining an FID of 107.47 even at 6 meters lane shift. Our code is available at https://github.com/wangyuanbiubiubiu/FaithFusion.