Deep learning models outperform ML for transferable satellite bathymetry

Researchers have compared machine learning and deep learning models for satellite-derived bathymetry (SDB), focusing on their ability to transfer knowledge across different geographical regions. The study found that deep learning models, particularly CNNs like ResNet and ConvNeXt, demonstrated superior performance and transferability compared to traditional Random Forest models. Key improvements included optimizing training data continuity and using a weighted RMSE loss function, which significantly reduced errors, especially in shallower waters. AI

IMPACT Deep learning models show promise for more robust and scalable satellite-derived bathymetry, potentially improving coastal mapping accuracy across diverse regions.

RANK_REASON This is a research paper presenting a comparative assessment of machine learning and deep learning models for a specific application. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• ConvNeXt-Large
• EfficientNet-B4
• Great Barrier Reef
• Joachim Moortgat
• MagicBathyNet
• Pratas Island
• Random Forest
• ResNet-101
• ResNet-50
• Sentinel-2
• U-Net

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →