This project uses cutting-edge satellite technology to monitor beluga whales and harp seals in the Canadian Arctic. Satellite images can capture wildlife on a large scale, but automatically detecting small animals in these images is challenging because existing methods often miss uncertain or ambiguous whale appearances.

We developed an intelligent system that combines advanced image annotation techniques (Segment Anything Model) with AI detection technology (YOLOv8) to automatically identify whales and seals in satellite imagery. The innovation lies in training the system not only on clearly visible whales, but also on uncertain cases that humans might overlook, making the detection more comprehensive and practical for real-world conservation efforts. Our system achieved 72.2% accuracy for whale detection and 70.3% for seal detection, even in crowded scenes. This approach significantly reduces the time and effort needed for manual annotation and can be adapted to monitor other species and habitats, ultimately supporting marine conservation and ecological research.

• Back