Why It Matters
Accurately tracking individual animals without invasive methods transforms ecological research, enabling better conservation and management of keystone species like brown bears. As climate change and human activity reshape habitats, scalable AI tools like PoseSwin provide timely, cost‑effective ways to monitor population dynamics across vast, remote landscapes.
Summary
In this episode, researchers from EPFL and Alaska Pacific University discuss PoseSwin, an AI system that identifies individual brown bears in Alaska despite seasonal changes in weight and coat. By focusing on stable head features and incorporating pose-aware transformer models, PoseSwin outperforms body‑only approaches and can even flag previously unseen bears. The team demonstrated the model’s utility on citizen‑science photos from Katmai National Park, showing its potential for large‑scale, non‑invasive wildlife monitoring. Open‑source release of the algorithm and data promises broader application to other species, from macaques to chimpanzees.
AI enables a Who’s Who of brown bears in Alaska
Comments
Want to join the conversation?
Loading comments...