deep learning
deep learning on World Data Ocean: a running collection of 5 stories we have gathered and hand-picked because they are worth your time. Every post here touches on deep learning in some way — the news, the analysis, the deep dives, and the occasional surprise find. A Global Hub for Ocean Intelligence 4 World Data Ocean is a centralized digital platform where researchers, scientists, and ocean enthusiasts converge to explore, analyze, and… New stories are added to this page as we find them, so check back if you want to keep up with what is happening around deep learning, or subscribe to the RSS feed to get them as soon as they are published. Browse the collection below, or head back to the homepage to see everything World Data Ocean is covering right now.
Using semi-supervised learning to detect beluga whales from aerial image sequences
Precise beluga whale population monitoring is essential for Arctic conservation, yet manual annotation of aerial imagery presents significant logistical and cost barriers. This study investigates a solution: semi-supervised learning. We systematically evaluated SEMI-DETR, a novel detection transformer, against supervised methods, demonstrating a 20% mean Average Precision improvement with just 1% of labeled data. Notably, calf detection—critical for assessing reproductive health—benefited most. These findings establish an empirical benchmark and offer practical guidance for resource-constrained conservation programs, mirroring approaches explored in our related work on plankton monitoring.
Plankton imager 10 monitoring in the southern North Sea: an open workflow for classification, morphometry and DwC-A publication
Plankton, foundational to marine food webs, are increasingly vulnerable to climate change impacts. High-frequency plankton imaging offers crucial insights into community composition and size, yet standardized workflows for processing and publishing these data remain scarce. This study details the deployment of the Plankton Imager 10 (Pi-10) in the southern North Sea and introduces an open, modular pipeline for data processing, classification, and Darwin-Core Archive (DwC-A) publication.
Range estimation of low-frequency underwater acoustic target based on deep learning architecture with data augmentation
Accurate underwater acoustic ranging is critical for numerous applications, yet traditional methods often struggle with precision and adaptability. This study introduces a novel approach utilizing a ResNet-UNet architecture enhanced by data augmentation to estimate the range of low-frequency underwater targets. Leveraging covariance matrix data, the method significantly improves accuracy, notably outperforming matched field processing and conventional neural networks, even under challenging signal conditions. Experimental validation using SWellEX-96 data demonstrates the efficacy of this integrated data ecosystem.
Multi-horizon forecasting of coastal high water temperature events for operational marine-heatwave early warning in South Korean aquaculture: a walk-forward benchmark of eleven statistical and deep-learning models
Coastal aquaculture in South Korea faces repeated losses due to marine-heatwave events impacting key species like olive flounder. Addressing this, a recent study benchmarks eleven statistical and deep-learning models for forecasting high water temperatures, culminating in the development of the Hybrid GNN-BiLSTM. This innovative architecture demonstrates superior long-horizon predictive skill, achieving the lowest root-mean-square error and calibrated probabilistic forecasts. The research highlights a 38-hour advisory lead time, demonstrating the potential for operationally actionable early warnings—a critical advancement for protecting vulnerable aquaculture operations.
Machine learning, eDNA and citizen science in monitoring and assessing biodiversity and invasive alien species at sea
Emerging technologies are revolutionizing marine biodiversity and invasive alien species monitoring. This research, presented at a recent summer school, details the integration of machine learning, citizen science, and environmental DNA (eDNA) alongside remote sensing and drones. Machine learning, utilizing frameworks like Essential Biodiversity Variables (EBVs), enables efficient species identification and global data synthesis. Initiatives like iNaturalist exemplify the power of citizen science combined with AI validation.