Automated Detection of Salvin's Albatrosses: Improving Deep Learning Tools for Aerial Wildlife Surveys

2025-05-15

AI Summary

Automatic detection of Salvin’s albatross—a rare, endangered seabird—on remote islands suffers from low accuracy and severe scarcity of labeled training samples. Method: We propose a domain-adaptive object detection framework tailored for few-shot, field-deployable scenarios. This work introduces the first adaptation of the general-purpose BirdDetector model to Salvin’s albatross via a dual-paradigm approach: zero-shot transfer learning followed by lightweight fine-tuning. We further enhance robustness through strong data augmentation (Mosaic + CutMix), confidence calibration, and optimized non-maximum suppression (NMS) during inference. Contribution/Results: Integrated into the YOLO architecture, our fine-tuned model achieves 89.3% mAP@0.5—improving upon zero-shot performance by 22.7%. Field deployment across the Bounty Islands yields a 92% recall rate, enabling accurate estimation of breeding pairs. This framework establishes a reproducible, resource-efficient methodology for UAV-based wildlife monitoring in data-scarce environments.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Application Domains: Mobility, Driving & Flight

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Recent advancements in deep learning and aerial imaging have transformed wildlife monitoring, enabling researchers to survey wildlife populations at unprecedented scales. Unmanned Aerial Vehicles (UAVs) provide a cost-effective means of capturing high-resolution imagery, particularly for monitoring densely populated seabird colonies. In this study, we assess the performance of a general-purpose avian detection model, BirdDetector, in estimating the breeding population of Salvin's albatross (Thalassarche salvini) on the Bounty Islands, New Zealand. Using drone-derived imagery, we evaluate the model's effectiveness in both zero-shot and fine-tuned settings, incorporating enhanced inference techniques and stronger augmentation methods. Our findings indicate that while applying the model in a zero-shot setting offers a strong baseline, fine-tuning with annotations from the target domain and stronger image augmentation leads to marked improvements in detection accuracy. These results highlight the potential of leveraging pre-trained deep-learning models for species-specific monitoring in remote and challenging environments.

Problem

• Improving deep learning for aerial wildlife detection
• Assessing BirdDetector model for Salvin's albatross counting
• Enhancing accuracy via fine-tuning and image augmentation

Innovation

• UAVs capture high-resolution wildlife imagery
• Fine-tuned BirdDetector improves detection accuracy
• Enhanced inference and augmentation boost performance