Manually Annotated Drone Imagery Dataset for Automatic Coastline Delineation

Abstract Automatic delineation of coastline in coastal zones is an essential task for various applications including protection of coastal regions, disaster management, and planning. The lack of availability of manually annotated high resolution datasets tailored for AI in coastal research remains a...

Full description

Saved in:
Bibliographic Details
Main Authors: Kamran Tanwari, Paweł Terefenko, Jakub Śledziowski, Andrzej Giza
Format: Article
Language:English
Published: Nature Portfolio 2025-03-01
Series:Scientific Data
Online Access:https://doi.org/10.1038/s41597-025-04685-7
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Abstract Automatic delineation of coastline in coastal zones is an essential task for various applications including protection of coastal regions, disaster management, and planning. The lack of availability of manually annotated high resolution datasets tailored for AI in coastal research remains a concern. Therefore, we created an open source, UAV captured and high resolution RGB dataset named MADRID (Manually Annotated DRone Imagery Dataset). It was recorded during six-separate UAV fly-over flight paths in two different types of coasts in Poland, Miedzyzdroje - cliff coast and in Mrzezyno - dune coast. The dataset comprises of 3691 high-resolution images and each image is accompanied by manually provided coastline annotations utilizing novel polyline annotation technique for the first time. The data is pre-split into training and test data subsets suitable for semantic segmentation tasks. The output of this study has implications for protection of coastal regions and marine ecosystems, delivering valuable insights into the overall situation in the Southern Baltic Sea. All data is organized and structured to comply with FAIR principles.
ISSN:2052-4463