# Features of interest from a multi-season satellite survey of baleen whales on the West Antarctic Peninsula

## TL;DR

This paper introduces a dataset of annotated satellite images of baleen whales in Antarctica to help train automated wildlife detection systems.

## Contribution

The novel contribution is a manually annotated dataset of 819 whale-related features from satellite imagery to support automated detection model training.

## Key findings

- A dataset of 819 annotated Features of Interest (FOIs) was created from satellite imagery of Wilhelmina Bay.
- The dataset was annotated by seven observers across ~1,900 km2 of imagery from 2018 to 2022.
- The dataset aims to expedite the development of automated systems for wildlife monitoring in remote regions.

## Abstract

The application of very high-resolution satellite imagery for the purpose of studying wildlife, particularly in remote regions, has gained significant traction in recent years. With this, there has been an exponential increase in the volume of satellite data collected, which has fostered a shift towards the use of automated systems to increase processing efficiency. However, these automated systems require manually annotated data on which to be trained, which is lacking due to the time required to manually annotate satellite imagery and the lack of published records to collaboratively build large enough training datasets. Here, we present a dataset that describes a total of 819 annotated and classified Features of Interest (FOIs) from a multi-season baleen whale-focussed survey of Wilhelmina Bay on the Western Antarctic Peninsula. These data are comprised of FOIs that have been annotated and classified based on existing protocols by seven individual observers who scanned ~1,900 km2 of WorldView-3 imagery acquired between 2018 and 2022 to expedite the creation of training datasets for automated detection models.