Always Clear Depth: Robust Monocular Depth Estimation under Adverse Weather

2025/05/18 by Kui Jiang, Jiang, Kui, Jing Cao +7 · 2 citations
Computer Science · #Advanced Vision and Imaging #Artificial Intelligence (cs.AI) #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Image Enhancement Techniques

paper · pdf · doi:10.48550/arxiv.2505.12199

openalex publication_date 2025/05/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Monocular depth estimation is critical for applications such as autonomous driving and scene reconstruction. While existing methods perform well under normal scenarios, their performance declines in adverse weather, due to challenging domain shifts and difficulties in extracting scene information. To address this issue, we present a robust monocular depth estimation method called ACDepth from the perspective of high-quality training data generation and domain adaptation. Specifically, we introduce a one-step diffusion model for generating samples that simulate adverse weather conditions, constructing a multi-tuple degradation dataset during training. To ensure the quality of the generated degradation samples, we employ LoRA adapters to fine-tune the generation weights of diffusion model. Additionally, we integrate circular consistency loss and adversarial training to guarantee the fidelity and naturalness of the scene contents. Furthermore, we elaborate on a multi-granularity knowledge distillation strategy (MKD) that encourages the student network to absorb knowledge from both the teacher model and pretrained Depth Anything V2. This strategy guides the student model in learning degradation-agnostic scene information from various degradation inputs. In particular, we introduce an ordinal guidance distillation mechanism (OGD) that encourages the network to focus on uncertain regions through differential ranking, leading to a more precise depth estimation. Experimental results demonstrate that our ACDepth surpasses md4all-DD by 2.50% for night scene and 2.61% for rainy scene on the nuScenes dataset in terms of the absRel metric.

Citations

• Diffusion Models for Monocular Depth Estimation: Overcoming Challenging Conditions
• Depth Anything V2
• Self-Supervised Monocular Depth Estimation in the Dark: Towards Data Distribution Compensation
• One-Step Image Translation with Text-to-Image Models
• Adversarial Diffusion Distillation
• Self-supervised Monocular Depth Estimation: Let's Talk About The Weather
• ZoeDepth: Zero-shot Transfer by Combining Relative and Metric Depth
• T2I-Adapter: Learning Adapters to Dig out More Controllable Ability for Text-to-Image Diffusion Models
• EVEN: An Event-Based Framework for Monocular Depth Estimation at Adverse Night Conditions
• STEPS: Joint Self-supervised Nighttime Image Enhancement and Depth Estimation
• Lite-Mono: A Lightweight CNN and Transformer Architecture for Self-Supervised Monocular Depth Estimation
• SC-DepthV3: Robust Self-supervised Monocular Depth Estimation for Dynamic Scenes
• Towards Accurate Reconstruction of 3D Scene Shape from A Single Monocular Image
• When the Sun Goes Down: Repairing Photometric Losses for All-Day Depth Estimation
• Learning Monocular Depth Estimation via Selective Distillation of Stereo Knowledge
• Self-supervised Monocular Depth Estimation for All Day Images using Domain Separation
• Regularizing Nighttime Weirdness: Efficient Self-supervised Monocular Depth Estimation in the Dark
• LoRA: Low-Rank Adaptation of Large Language Models
• Vision Transformers for Dense Prediction
• HR-Depth: High Resolution Self-Supervised Monocular Depth Estimation
• Toward Hierarchical Self-Supervised Monocular Absolute Depth Estimation for Autonomous Driving Applications
• DeFeat-Net: General Monocular Depth via Simultaneous Unsupervised Representation Learning
• Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement
• Unsupervised Scale-consistent Depth and Ego-motion Learning from Monocular Video
• Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer
• Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-Shot Cross-Dataset Transfer
• 3D Packing for Self-Supervised Monocular Depth Estimation
• nuScenes: A multimodal dataset for autonomous driving
• BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning
• BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning
• Semantic Foggy Scene Understanding with Synthetic Data
• Unsupervised Learning of Depth and Ego-Motion from Video
• 1 year, 1000 km: The Oxford RobotCar dataset
• Unsupervised Monocular Depth Estimation with Left-Right Consistency
• Unsupervised CNN for Single View Depth Estimation: Geometry to the Rescue
• Deep Residual Learning for Image Recognition
• Distilling the Knowledge in a Neural Network
• Deep Convolutional Neural Fields for Depth Estimation from a Single Image
• EC-Depth: Exploring the consistency of self-supervised monocular depth estimation in challenging scenes

Cited by

• DASP: Self-supervised Nighttime Monocular Depth Estimation with Domain Adaptation of Spatiotemporal Priors
• Learning Depth from Past Selves: Self-Evolution Contrast for Robust Depth Estimation

Related

• Stealing Stable Diffusion Prior for Robust Monocular Depth Estimation
• Self-supervised Monocular Depth Estimation: Let's Talk About The Weather
• EC-Depth: Exploring the consistency of self-supervised monocular depth estimation in challenging scenes
• Synthetic-to-Real Self-supervised Robust Depth Estimation via Learning with Motion and Structure Priors
• ER-LoRA: Effective-Rank Guided Adaptation for Weather-Generalized Depth Estimation