Seabed-Net: A multi-task network for joint bathymetry estimation and seabed classification from remote sensing imagery in shallow waters

2025-10-22

AI Summary

Conventional approaches to bathymetric inversion and seabed classification in shallow waters have been modeled separately, hindering joint optimization and limiting performance. Method: This paper proposes the first multi-task deep learning framework jointly performing bathymetric regression and pixel-wise seabed classification. It introduces a dual-branch encoder with an attention-based feature fusion module, pioneering the integration of Swin-Transformer into remote sensing–based collaborative seafloor inversion to enable cross-task feature interaction; additionally, a dynamic uncertainty weighting scheme adaptively balances multi-task losses. The framework supports multi-resolution remote sensing inputs. Results: Evaluated on two coastal field sites, it achieves 10–30% reduction in bathymetric RMSE and up to 8% improvement in overall seabed classification accuracy, with sharper class boundaries and significantly reduced errors in low-contrast regions—outperforming both traditional models and state-of-the-art methods across all metrics.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Search and Optimization: Learning to Search
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• 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
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Accurate, detailed, and regularly updated bathymetry, coupled with complex semantic content, is essential for under-mapped shallow-water environments facing increasing climatological and anthropogenic pressures. However, existing approaches that derive either depth or seabed classes from remote sensing imagery treat these tasks in isolation, forfeiting the mutual benefits of their interaction and hindering the broader adoption of deep learning methods. To address these limitations, we introduce Seabed-Net, a unified multi-task framework that simultaneously predicts bathymetry and pixel-based seabed classification from remote sensing imagery of various resolutions. Seabed-Net employs dual-branch encoders for bathymetry estimation and pixel-based seabed classification, integrates cross-task features via an Attention Feature Fusion module and a windowed Swin-Transformer fusion block, and balances objectives through dynamic task uncertainty weighting. In extensive evaluations at two heterogeneous coastal sites, it consistently outperforms traditional empirical models and traditional machine learning regression methods, achieving up to 75% lower RMSE. It also reduces bathymetric RMSE by 10-30% compared to state-of-the-art single-task and multi-task baselines and improves seabed classification accuracy up to 8%. Qualitative analyses further demonstrate enhanced spatial consistency, sharper habitat boundaries, and corrected depth biases in low-contrast regions. These results confirm that jointly modeling depth with both substrate and seabed habitats yields synergistic gains, offering a robust, open solution for integrated shallow-water mapping. Code and pretrained weights are available at https://github.com/pagraf/Seabed-Net.

Problem

• Simultaneously predicts bathymetry and seabed classification from imagery
• Addresses limitations of isolated approaches for depth and seabed mapping
• Enhances accuracy in shallow-water environments through joint modeling

Innovation

• Multi-task network for bathymetry and seabed classification
• Dual-branch encoders with cross-task feature fusion
• Dynamic task uncertainty weighting balances objectives