DASP: Self-supervised Nighttime Monocular Depth Estimation with Domain Adaptation of Spatiotemporal Priors

2025-12-16

AI Summary

Nighttime self-supervised monocular depth estimation suffers severe performance degradation due to low illumination, texture scarcity, and motion blur. To address this, we propose DASP, a domain-adaptive framework. Methodologically, DASP introduces the SPLB module—integrating motion-aware orthogonal differential temporal modeling with axial spatial attention—and enforces inter-frame geometric consistency via a novel 3D-consistent projection loss operating in a shared 3D space. Technically, it synergistically combines adversarial learning, local asymmetric convolution, and global axial attention to enhance depth robustness in weak-texture and dynamic regions. Evaluated on the Oxford RobotCar and nuScenes nighttime benchmarks, DASP achieves state-of-the-art performance, significantly improving depth accuracy in texture-deficient areas and motion-blurred regions.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Self-supervised monocular depth estimation has achieved notable success under daytime conditions. However, its performance deteriorates markedly at night due to low visibility and varying illumination, e.g., insufficient light causes textureless areas, and moving objects bring blurry regions. To this end, we propose a self-supervised framework named DASP that leverages spatiotemporal priors for nighttime depth estimation. Specifically, DASP consists of an adversarial branch for extracting spatiotemporal priors and a self-supervised branch for learning. In the adversarial branch, we first design an adversarial network where the discriminator is composed of four devised spatiotemporal priors learning blocks (SPLB) to exploit the daytime priors. In particular, the SPLB contains a spatial-based temporal learning module (STLM) that uses orthogonal differencing to extract motion-related variations along the time axis and an axial spatial learning module (ASLM) that adopts local asymmetric convolutions with global axial attention to capture the multiscale structural information. By combining STLM and ASLM, our model can acquire sufficient spatiotemporal features to restore textureless areas and estimate the blurry regions caused by dynamic objects. In the self-supervised branch, we propose a 3D consistency projection loss to bilaterally project the target frame and source frame into a shared 3D space, and calculate the 3D discrepancy between the two projected frames as a loss to optimize the 3D structural consistency and daytime priors. Extensive experiments on the Oxford RobotCar and nuScenes datasets demonstrate that our approach achieves state-of-the-art performance for nighttime depth estimation. Ablation studies further validate the effectiveness of each component.

Problem

• Addresses nighttime depth estimation challenges due to low visibility and illumination
• Leverages spatiotemporal priors to restore textureless areas and handle blurry regions
• Uses adversarial and self-supervised branches for domain adaptation from daytime data

Innovation

• Adversarial network with spatiotemporal priors learning blocks
• Spatial-based temporal and axial spatial learning modules
• 3D consistency projection loss for structural optimization