WaterSplat-SLAM: Photorealistic Monocular SLAM in Underwater Environment

2026-04-06

AI Summary

This work addresses the poor robustness of existing monocular underwater SLAM methods in complex environments and their inability to generate high-fidelity, photorealistic dense maps. To overcome these limitations, we propose WaterSplat-SLAM, which introduces, for the first time, a semantic medium-aware mechanism into monocular underwater SLAM. Our approach leverages semantic medium-aware filtering within a two-view 3D reconstruction framework to achieve robust camera tracking and depth estimation. Furthermore, by integrating semantic-guided rendering with an online medium-aware Gaussian map representation, the system produces compact yet visually realistic dense reconstructions. Extensive evaluation on multiple underwater datasets demonstrates that WaterSplat-SLAM significantly improves both tracking robustness and the photometric realism of the generated maps.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Computer Vision: Motion & Tracking
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Underwater monocular SLAM is a challenging problem with applications from autonomous underwater vehicles to marine archaeology. However, existing underwater SLAM methods struggle to produce maps with high-fidelity rendering. In this paper, we propose WaterSplat-SLAM, a novel monocular underwater SLAM system that achieves robust pose estimation and photorealistic dense mapping. Specifically, we couple semantic medium filtering into two-view 3D reconstruction prior to enable underwater-adapted camera tracking and depth estimation. Furthermore, we present a semantic-guided rendering and adaptive map management strategy with an online medium-aware Gaussian map, modeling underwater environment in a photorealistic and compact manner. Experiments on multiple underwater datasets demonstrate that WaterSplat-SLAM achieves robust camera tracking and high-fidelity rendering in underwater environments.

Problem

• underwater SLAM
• monocular SLAM
• photorealistic mapping
• dense mapping
• high-fidelity rendering

Innovation

• Underwater SLAM
• Photorealistic Rendering
• Semantic-guided Reconstruction
• Gaussian Splatting
• Monocular Vision