This web application uses cookies and other tracking technologies to ensure you get the best experience.
SC/69B/PH/08
Download this file1.1 MB |
Download |
View directly in browser1.1 MB |
View in browser |
Resource ID
22196
Access
Open
Document Number
SC/69B/PH/08
Full Title
Deep learning approach for automated photo identification using vertical aerial drone imagery of fin whales (Balaenoptera physalus)
Author
Kilian Huss, Alexander Rychwalski, Sacha Viquerat, Helena Herr
Authors Summary
This paper presents a conceptualisation for a deep learning-based framework for automated photo identiffication of Southern Hemisphere fin whales using aerial drone footage of characteristic dorsal pigmentation patterns. By focusing on the Central Chevron Pattern (CCP) and Blaze, the framework aims to develop a novel aerial photo ID approach by leveraging a semi-supervised workflow and thereby reducing manual labelling efforts. The objective is to identify key feature regions with high inter-individual variability and to explore subsequent automation. Model predictions are then to be yielded by employing a deep convolutional neural network (CNN) architecture for (human) facial recognition to ensure robust individual discrimination.
Publisher
IWC
Publication Year
2024