AutoWeather4D: Autonomous Driving Video Weather Conversion via G-Buffer Dual-Pass Editing

View PDF HTML (experimental)

Abstract:Generative video models have significantly advanced the photorealistic synthesis of adverse weather for autonomous driving; however, they consistently demand massive datasets to learn rare weather scenarios. While 3D-aware editing methods alleviate these data constraints by augmenting existing video footage, they are fundamentally bottlenecked by costly per-scene optimization and suffer from inherent geometric and illumination entanglement. In this work, we introduce AutoWeather4D, a feed-forward 3D-aware weather editing framework designed to explicitly decouple geometry and illumination. At the core of our approach is a G-buffer Dual-pass Editing mechanism. The Geometry Pass leverages explicit structural foundations to enable surface-anchored physical interactions, while the Light Pass analytically resolves light transport, accumulating the contributions of local illuminants into the global illumination to enable dynamic 3D local relighting. Extensive experiments demonstrate that AutoWeather4D achieves comparable photorealism and structural consistency to generative baselines while enabling fine-grained parametric physical control, serving as a practical data engine for autonomous driving.

Comments: Project Page: this https URL | Github: this https URL

Subjects:

Computer Vision and Pattern Recognition (cs.CV)

Cite as: arXiv:2603.26546 [cs.CV]

(or arXiv:2603.26546v3 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2603.26546

arXiv-issued DOI via DataCite

Submission history

From: Tianyu Liu [view email] [v1] Fri, 27 Mar 2026 15:56:47 UTC (27,825 KB) [v2] Tue, 31 Mar 2026 03:08:07 UTC (27,825 KB) [v3] Thu, 2 Apr 2026 02:56:53 UTC (27,825 KB)