TRENDING

TECHNOLOGY NEWS FROM AROUND THE WORLD

→
AI Enhances Beluga Whale Tracking via Aerial Drone Imagery
Phys.org BiologyIN-DEPTH NEWS

AI Enhances Beluga Whale Tracking via Aerial Drone Imagery

Researchers have developed a new AI approach using semi-supervised learning to detect and track beluga whales and their calves from high-resolution drone images.

·Phys.org Biology
Reading progress0%

Challenges in Arctic Marine Monitoring

For researchers working in the Arctic, counting beluga whales from aerial perspectives involves the grueling task of examining thousands of drone images. Identifying individual animals within these high-resolution sequences is difficult, particularly when distinguishing between different individuals. The difficulty intensifies when attempting to locate calves, which are characterized by their smaller size, lower population numbers, and highly variable appearances. Accurately identifying these calves is essential for understanding reproductive success, population growth trends, and the overall long-term health of beluga whale communities. Currently, the process relies heavily on labor-intensive manual annotation by human experts.

Advancing Detection with SEMI-DETR

To address these limitations, a collaborative team including Florida Atlantic University, Harbor Branch Oceanographic Institute, Fisheries and Oceans Canada, and the University of Manitoba has developed a deep learning approach. Published in Frontiers in Marine Science, the study utilized a dataset of 7,655 high-resolution aerial drone images. Researchers compared five different AI models to evaluate performance across various levels of human-labeled data. Their semi-supervised approach, known as SEMI-DETR, demonstrated remarkable efficiency. By using only 34 labeled images—representing just 1% of the available data—the model achieved a mean average precision of 19.2, outperforming the best supervised model by 20%.

Superior Performance in Calf Identification

The implementation of semi-supervised learning proved especially beneficial for detecting beluga calves, which are traditionally harder to spot than adults. At the 1% labeling threshold, the SEMI-DETR model improved calf detection by 24.3% compared to the supervised baseline. In contrast, the improvement for adult whales stood at 16.8%. This indicates that the AI can effectively learn from minimal expert guidance to identify complex, variable targets. Overall, the research suggests that comparable detection accuracy can be reached using five to ten times less manual annotation, significantly reducing the time and financial costs associated with preparing imagery for automated analysis.

Collaborative Scientific and Engineering Efforts

This research represents a significant intersection of engineering and environmental science. Senior author Hanqi Zhuang noted that artificial intelligence provides a new pathway for conservationists to interpret the massive volumes of imagery collected by modern monitoring tools. By allowing technology to learn from small amounts of expert input and then apply that knowledge to vast unlabeled datasets, the method becomes more scalable. Stella Batalama, dean of the College of Engineering and Computer Science, emphasized that such innovative technologies create practical solutions to strengthen conservation efforts and protect vulnerable marine ecosystems against complex environmental challenges.

Transforming Wildlife Observation Methods

The development of these computer vision tools has the potential to transform wildlife monitoring globally. Instead of requiring researchers to painstakingly label thousands of images before an AI system becomes functional, semi-supervised learning makes sophisticated tools more accessible to scientists working in challenging environments. This approach provides a vital benchmark for marine mammal monitoring, especially for conservation programs operating with limited resources for manual data annotation. The ability to automate tasks that previously required extensive human effort allows for more frequent and widespread monitoring of species in remote and difficult-to-reach Arctic locations.

Implications for Long-term Conservation

Beyond simple counting, the technology offers deeper insights into marine biology. Co-author Greg O'Corry-Crowe highlighted that the ability to process far more observations than humanly possible allows scientists to better understand beluga movement patterns, habitat use, and reproductive success. Understanding how these patterns change over time is crucial for managing populations. Ultimately, this integration of artificial intelligence and engineering provides a robust framework for addressing environmental crises, offering a way to monitor how vulnerable species respond to changing conditions in the High Arctic through more efficient and scalable technological means.

Illustrative news photo · Unsplash / Unsplash ↗

Sources and verification

Be the first reader ✎ Write a comment

Comments

Reader question: What is your view or technical question about this story?

No comments yet. Be the first to contribute.

Write a comment

Your first and last name will appear publicly alongside your comment.

Sample comments and editor reply

Discussion example — not real reader feedback. Names and conversations in this area are fictional. Likes and replies here are not saved and do not count as real engagement.

Example counters: 16,679 reads · 799 likes — not actual analytics

Ayşe Yılmaz

Reading “AI Enhances Beluga Whale Tracking via Aerial Drone Imagery” raised this question for me: uncertainty matters more than a dramatic headline.

Sample editor reply

Editor note: Forward-looking claims in an announcement are not the same as established performance results. Further evidence may change the assessment.

Fatma Demir

It is important to distinguish what the source verifies from projections about the future.

Mehmet Kaya

The next version's measurements should be compared with the initial announcement.