Open-Weather Robust 3D Detection via Dual-Critic Diffusion Alignment

2026-07-02

AI Summary

This work addresses the significant performance degradation of existing 3D object detection methods under adverse weather conditions in open-world settings, which stems from their reliance on the closed-world assumption that training and testing data share aligned weather conditions. To overcome this limitation, the authors propose a weather-agnostic feature alignment framework that leverages a 4D radar-guided conditional diffusion model to recover LiDAR features degraded by adverse weather. The approach incorporates a dual-discriminator mechanism: a detection-guided discriminator preserves semantic fidelity, while a weather-adversarial discriminator aligns the feature distribution with that of clear-weather data. Notably, the method requires neither explicit weather modeling, paired data, nor weather labels, and achieves strong generalization to unseen weather types and intensities for the first time. Evaluated on a newly introduced open-weather benchmark, it substantially outperforms current state-of-the-art methods, significantly enhancing the robustness of 3D detection.

Abstract

Robust 3D object detection under adverse weather remains a critical hurdle for autonomous driving. Despite progress with LiDAR-4D radar fusion, most methods are constrained by a closed-world assumption, implicitly requiring training and test weather to align in both type and severity. This premise fails in practice: the open-ended nature of weather, and even variations within a single type like rain, cause dramatically different LiDAR degradation patterns, leading to significant performance drops in unseen conditions. To address this, we present Dual-Critic Guided Diffusion Alignment (DCDA), a weather-agnostic framework that learns to recover degraded LiDAR features toward a clean manifold. Rather than modeling specific weather types, DCDA employs a 4D radar-conditioned diffusion process to progressively refine features, guided by two complementary critics. (i) A detection-guided critic, anchored by a pre-trained clean-weather model, ensures that the refined features retain object-level discriminability and localization accuracy. (ii) A weather adversarial critic enforces holistic distributional consistency with clean-weather representations. By aligning features through semantic and distributional constraints rather than explicit weather modeling, DCDA generalizes effectively to unseen weather types and severities without requiring paired data or weather labels. We further introduce a structured open-weather benchmark with held-out type-severity combinations and extensive experiments verify DCDA's advantages.

Problem

• open-weather
• 3D object detection
• LiDAR degradation
• domain generalization
• autonomous driving

Innovation

• diffusion alignment
• weather-agnostic perception
• 4D radar-LiDAR fusion
• adversarial critic
• open-world 3D detection