A 1.2 Billion Pixel Human-Labeled Dataset for Data-Driven Classification of Coastal Environments

Abstract

The world’s coastlines are spatially highly variable, coupled-human-natural systems that comprise a nested hierarchy of component landforms, ecosystems, and human interventions, each interacting over a range of space and time scales. Understanding and predicting coastline dynamics necessitates frequent observation from imaging sensors on remote sensing platforms. Machine Learning models that carry out supervised (i.e., human-guided) pixel-based classification, or image segmentation, have transformative applications in spatio-temporal mapping of dynamic environments, including transient coastal landforms, sediments, habitats, waterbodies, and water flows. However, these models require large and well-documented training and testing datasets consisting of labeled imagery. We describe “Coast Train,” a multi-labeler dataset of orthomosaic and satellite images of coastal environments and corresponding labels. These data include imagery that are diverse in space and time, and contain 1.2 billion labeled pixels, representing over 3.6 million hectares. We use a human-in-the-loop tool especially designed for rapid and reproducible Earth surface image segmentation. Our approach permits image labeling by multiple labelers, in turn enabling quantification of pixel-level agreement over individual and collections of images.

Fully convolutional neural networks applied to large-scale marine morphology mapping

“…While interest in DL has been shown early on by the marine community for ecological and habitat mapping (Gazis et al, 2018;Yasir et al, 2021), only a few studies have been focused on automated identification with DL of seabed geomorphological features or textures (McClinton et al, 2012;Valentine et al, 2013;Juliani, 2019;Keohane and White, 2022;Lundine et al, 2023), even though the significance of geomorphology for habitat distribution is widely acknowledged (Brown et al, 2011;Lecours et al, 2016;Harris and Baker, 2020). Deep Learning in geomorphology has found instead a more fertile ground in coastal and geohazard studies (Ma and Mei, 2021;Buscombe et al, 2023), and in outer space, in particular for Martian or Lunar geology, where several studies have taken advantage of the high resolution optical imagery available and attempted to separate specific landforms from a background (Foroutan and Zimbelman, 2017;Palafox et al, 2017;Wang et al, 2017;Rubanenko et al, 2021), or more generally characterise the ground surface to identify optimal landing spots or assess rover traversability (Wilhelm et al, 2020;Barrett et al, 2022). Barrett et al (2022) in particular have demonstrated the potential of large-scale exploratory morphological mapping, where machine learning assists the geomorphologist to isolate sections of interest in the dataset, sifting through an enormous dataset.…”

Observing coastal wetland transitions using national land cover products

“…Another challenge for semantic mapping with deep convolutional neural networks is the need for large amounts of training data. In response, expansive libraries of training data (i.e., labeled images and points) have been built that can be used for training and tuning deep convolutional neural networks or other algorithms (Murray et al, 2022; Buscombe et al, 2023).…”

Introducing ICEDAP: An ‘Iterative Coastal Embayment Delineation and Analysis Process’ with Applications for the Management of Coastal Change

“…This includes coastline migration datasets providing temporal accounts of both past and future rates of erosion/accretion [75]. Global inventories of coastal habitats have been widely developed; these identify specific habitat types and track their changes, both for habitats overall and for specific habitats like salt marshes or mudflats [76][77][78]. Further, datasets for human development within embayed systems (e.g., land use datasets and inventories of infrastructure or other human activities) are common on local and regional scales in select geographies (but often lack global coverage).…”