WeatherEdit: Controllable Weather Editing with 4D Gaussian Field

TL;DR

WeatherEdit is a comprehensive system that enables controllable, realistic weather effects in 3D scenes by combining diffusion-based background editing with a 4D Gaussian field for dynamic weather particles, suitable for autonomous driving simulations.

Contribution

It introduces a novel pipeline integrating diffusion models and a 4D Gaussian field for realistic, controllable weather effects in 3D scenes, addressing previous limitations in weather editing.

Findings

Able to generate diverse weather effects with adjustable severity.

Ensures consistent multi-view and multi-frame weather editing.

Demonstrates effectiveness on multiple driving datasets.

Abstract

In this work, we present WeatherEdit, a novel weather editing pipeline for generating realistic weather effects with controllable types and severity in 3D scenes. Our approach is structured into two key components: weather background editing and weather particle construction. For weather background editing, we introduce an all-in-one adapter that integrates multiple weather styles into a single pretrained diffusion model, enabling the generation of diverse weather effects in 2D image backgrounds. During inference, we design a Temporal-View (TV-) attention mechanism that follows a specific order to aggregate temporal and spatial information, ensuring consistent editing across multi-frame and multi-view images. To construct the weather particles, we first reconstruct a 3D scene using the edited images and then introduce a dynamic 4D Gaussian field to generate snowflakes, raindrops and fog in the scene. The attributes and dynamics of these particles are precisely controlled through physical-based modelling and simulation, ensuring realistic weather representation and flexible severity adjustments. Finally, we integrate the 4D Gaussian field with the 3D scene to render consistent and highly realistic weather effects. Experiments on multiple driving datasets demonstrate that WeatherEdit can generate diverse weather effects with controllable condition severity, highlighting its potential for autonomous driving simulation in adverse weather. See project page: https://jumponthemoon.github.io/w-edit