Improving Wildlife Management with AI: Species Detection and Classification from Camera Trap Data
Year: 2024
Type: article
Abstract: In this study, we explore advanced computer vision techniques to enhance wildlife management through the automatic detection and classification of animal species from camera trap images. Leveraging deep learning methods, our research focuses on the automated extraction of critical information from these images to support forest and wildlife management, biodiversity monitoring, and reintroduction program evaluations. We present a specialized data set with manually labeled and validated images and comprehensive metadata, including species identification, sex, age class, and unique IDs for individual animals. Our approach integrates both single-stage and two-stage detection and classification strategies, utilizing models such as YOLO and EfficientNet. Initial results demonstrate the effectiveness of our methods, achieving significant accuracies (up to 95%) and providing a user-friendly interface for further refinement of classifications. Future work will expand the data set and explore transformer-based deep neural networks to enhance the robustness and applicability of our wildlife classification system.
Authors: Mustafic, Sead, Dachs, Dominik, Prüller, Rainer, Schöggl, Florian, Perko, Roland
Institution: Joanneum Research
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
SDG: Life on land
Open Access status: green
OpenAlex ID: W6893106375
FWCI:
Cites:
Cited by:
Related to: