Dense geometry supervision for underwater depth estimation

2025-02-20

International Conference on Automata and Formal Languages

AI Summary

Monocular underwater depth estimation suffers from low accuracy and poor generalization due to severe image degradation and scarcity of labeled training data. To address this, we introduce FLSea—the first low-cost, multi-view synthesized underwater dataset featuring enhanced images and dense geometric supervision—and propose a physics-informed texture-depth fusion module. Our approach establishes a dedicated dense geometric supervision paradigm for underwater scenes and designs a physical-prior-guided feature fusion network to effectively mitigate the loss of depth cues induced by optical degradation. We further adopt a joint training strategy integrating multi-view depth estimation, underwater image enhancement, and self-supervised/semi-supervised learning. Evaluated on FLSea, our method reduces mean absolute error by 19.3% over baseline methods, significantly improving both accuracy and robustness. This work provides a reliable monocular depth estimation solution for autonomous underwater vehicles and marine exploration.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The field of monocular depth estimation is continually evolving with the advent of numerous innovative models and extensions. However, research on monocular depth estimation methods specifically for underwater scenes remains limited, compounded by a scarcity of relevant data and methodological support. This paper proposes a novel approach to address the existing challenges in current monocular depth estimation methods for underwater environments. We construct an economically efficient dataset suitable for underwater scenarios by employing multi-view depth estimation to generate supervisory signals and corresponding enhanced underwater images. we introduces a texture-depth fusion module, designed according to the underwater optical imaging principles, which aims to effectively exploit and integrate depth information from texture cues. Experimental results on the FLSea dataset demonstrate that our approach significantly improves the accuracy and adaptability of models in underwater settings. This work offers a cost-effective solution for monocular underwater depth estimation and holds considerable promise for practical applications.

Problem

• Addressing limited underwater monocular depth estimation methods
• Creating cost-effective dataset for underwater depth supervision
• Improving accuracy with texture-depth fusion for underwater scenes

Innovation

• Multi-view depth estimation for dataset creation
• Texture-depth fusion module for underwater imaging
• Cost-effective solution for underwater depth estimation