We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
AqUavplant Dataset: A High-Resolution Aquatic Plant Classification and Segmentation Image Dataset Using UAV
Summary
Researchers created a high-resolution image dataset of 31 aquatic plant species in Bangladesh using drone photography, designed to help train machine learning models for automated plant mapping. The dataset includes detailed segmentation masks that can help identify individual species, track plant growth, and monitor the spread of invasive species. This tool could support conservation efforts by making it easier to monitor aquatic plant biodiversity across large areas.
Aquatic vegetation species are declining gradually, posing a threat to the stability of aquatic ecosystems. The decline can be controlled with proper monitoring and mapping of the species for effective conservation and management. The Unmanned Ariel Vehicle (UAV) aka Drone can be deployed to comprehensively capture large area of water bodies for effective mapping and monitoring. This study developed the AqUavplant dataset consisting of 197 high resolution (3840px × 2160px, 4K) images of 31 aquatic plant species collected from nine different sites in Bangladesh. The DJI Mavic 3 Pro triple-camera professional drone is used with a ground sampling distance (GSD) value of 0.04-0.05 cm/px for optimal image collection without losing detail. The dataset is complemented with binary and multiclass semantic segmentation mask to facilitate ML based model development for automatic plant mapping. The dataset can be used to detect the diversity of indigenous and invasive species, monitor plant growth and diseases, measure the growth ratio to preserve biodiversity, and prevent extinction.
Sign in to start a discussion.
More Papers Like This
Data Study Group Final Report: Centre for Environment, Fisheries and Aquaculture Science
Machine learning was applied to the challenge of automatically classifying plankton species from underwater images collected by fisheries monitoring systems. The AI classifier could identify dozens of plankton categories with high accuracy, reducing the need for time-consuming manual identification. Automated plankton monitoring improves understanding of marine food web health and ecosystem responses to environmental change.
Plankton classification with high-throughput submersible holographic microscopy and transfer learning
Researchers used underwater holographic microscopes and transfer learning — an AI technique that applies knowledge from one task to another — to automatically classify diverse plankton species from images, including rare forms. The system shows promise for large-scale, automated ocean monitoring without needing constant human analysis.
Machine learning based approaches for prompt diagnosis of aquatic plant ailments
Researchers applied Deep Belief Networks and Isolation Forests to diagnose aquatic plant diseases from observational data, achieving an average DBN accuracy of 86% and an Isolation Forest true positive rate of 91%. The study demonstrated that machine learning can improve early detection of aquatic plant ailments, which is relevant to monitoring aquatic ecosystem health in microplastic-contaminated environments.
SNOWED: Automatically Constructed Dataset of Satellite Imagery for Water Edge Measurements
Researchers developed SNOWED, an automatically constructed dataset of satellite imagery with labeled water edges, enabling deep learning models to accurately detect and monitor shoreline changes for environmental monitoring applications.
Mini Uav-based Litter Detection on River Banks
Researchers developed a drone-based litter detection system combining high-resolution mapping, deep learning object detection, and vision-based localization that locates riverbank litter with decimeter-level accuracy, enabling automated monitoring of plastic pollution in urban waterway areas.