An annotated aerial imagery dataset for automated detection of harbour seals in Svalbard, Norway.

Arctic marine ecosystems are undergoing significant changes due to rapid climate warming, affecting species abundances and distributions. Harbour seals (Phoca vitulina), the most widely distributed pinniped in the world, are expanding their range northward in Arctic waters. Monitoring their ongoing distributional shift using traditional survey methods is logistically challenging. In particular, detecting and manually counting seals from aerial imagery is labour-intensive and costly. Integrating object detection algorithms into the analytical phase of surveys can greatly reduce post-processing time and costs. Here, we present 495 annotated images from the High Arctic Svalbard Archipelago with 7,085 harbour seal annotations highlighted as oriented bounding boxes. The dataset includes challenging background conditions with well-camouflaged and sparsely distributed animals, providing a valuable resource that complements existing datasets. While Svalbard is of particular ecological relevance as the fastest-warming region on the planet with advanced climate-driven range shifts, the detection methodology enabled by this dataset is applicable across the species' full geographic range. It supports the development of automated object detection models for ecological monitoring and population studies worldwide.