DriveFix: Spatio-Temporally Coherent Driving Scene Restoration

TL;DR

DriveFix introduces a spatio-temporally coherent multi-view restoration framework for driving scenes, leveraging diffusion transformers and geometry-aware training to improve 4D scene reconstruction and view synthesis in autonomous driving.

Contribution

It presents a novel interleaved diffusion transformer architecture that explicitly models temporal and spatial dependencies for coherent scene restoration.

Findings

Achieves state-of-the-art results on Waymo, nuScenes, and PandaSet datasets.

Reduces artifacts and improves texture consistency across views and time.

Enforces 3D geometric consistency in scene reconstruction.

Abstract

Recent advancements in 4D scene reconstruction, particularly those leveraging diffusion priors, have shown promise for novel view synthesis in autonomous driving. However, these methods often process frames independently or in a view-by-view manner, leading to a critical lack of spatio-temporal synergy. This results in spatial misalignment across cameras and temporal drift in sequences. We propose DriveFix, a novel multi-view restoration framework that ensures spatio-temporal coherence for driving scenes. Our approach employs an interleaved diffusion transformer architecture with specialized blocks to explicitly model both temporal dependencies and cross-camera spatial consistency. By conditioning the generation on historical context and integrating geometry-aware training losses, DriveFix enforces that the restored views adhere to a unified 3D geometry. This enables the consistent propagation of high-fidelity textures and significantly reduces artifacts. Extensive evaluations on the Waymo, nuScenes, and PandaSet datasets demonstrate that DriveFix achieves state-of-the-art performance in both reconstruction and novel view synthesis, marking a substantial step toward robust 4D world modeling for real-world deployment.