The Constant Eye: Benchmarking and Bridging Appearance Robustness in Autonomous Driving

TL;DR

This paper introduces navdream, a benchmark for testing appearance robustness in autonomous driving, and proposes a perception interface using a foundation model to improve zero-shot generalization under appearance shifts.

Contribution

The paper creates a high-fidelity benchmark isolating appearance effects and introduces a perception interface with a foundation model to enhance robustness without fine-tuning.

Findings

Existing algorithms degrade under appearance shifts.

The perception interface maintains performance across diverse models.

The benchmark isolates appearance effects from scene structure.

Abstract

Despite rapid progress, autonomous driving algorithms remain notoriously fragile under Out-of-Distribution (OOD) conditions. We identify a critical decoupling failure in current research: the lack of distinction between appearance-based shifts, such as weather and lighting, and structural scene changes. This leaves a fundamental question unanswered: Is the planner failing because of complex road geometry, or simply because it is raining? To resolve this, we establish navdream, a high-fidelity robustness benchmark leveraging generative pixel-aligned style transfer. By creating a visual stress test with negligible geometric deviation, we isolate the impact of appearance on driving performance. Our evaluation reveals that existing planning algorithms often show significant degradation under OOD appearance conditions, even when the underlying scene structure remains consistent. To bridge this gap, we propose a universal perception interface leveraging a frozen visual foundation model (DINOv3). By extracting appearance-invariant features as a stable interface for the planner, we achieve exceptional zero-shot generalization across diverse planning paradigms, including regression-based, diffusion-based, and scoring-based models. Our plug-and-play solution maintains consistent performance across extreme appearance shifts without requiring further fine-tuning. The benchmark and code will be made available.