Using AI to Identify Epaulette Sharks
Author Information
Author(s): Martina Lonati, Jahanbakht Mohammad, Atkins Danielle, Bierwagen Stacy L., Chin Andrew, Barnett Adam, Rummer Jodie L.
Primary Institution: James Cook University
Hypothesis
Can deep neural networks improve the reliability of photographic identification of epaulette sharks across different life stages?
Conclusion
The study found that while adult epaulette sharks can be reliably identified using AI with about 86% accuracy, immature life stages are not suitable for identification due to their changing patterns.
Supporting Evidence
- AI models can automate the identification process, reducing human error.
- Adult sharks maintained their identification patterns over time.
- Juvenile and neonate sharks showed unreliable patterns for identification.
Takeaway
Scientists used computers to help recognize different sharks by their unique patterns, but baby sharks are too different as they grow up, making it hard to tell them apart.
Methodology
The study developed an AI model to analyze photographs of epaulette sharks taken at different life stages and times to assess the reliability of their identification patterns.
Limitations
The study's findings may not apply to wild populations, and the model's effectiveness on immature sharks was limited due to their changing patterns.
Participant Demographics
The study involved a captive population of epaulette sharks, including eight adults, four juveniles, and five neonates.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website