Coral Reefs are one of the most diverse ecosystems on our planet. Due to the rise in global temperatures and various other threats, there has been over a 40% loss of coral reefs over the last 30 years alone. Monitoring of the health of coral reef ecosystems is critical for providing a deeper understanding of the ecology of these systems to form a scientific foundation for informing crucial management recommendations. However, monitoring of coral reef bleaching is currently a time-consuming, expensive, and laborious process.
In this project, Anne Lee, Shreya Ravi, and I curated a novel dataset of coral images in collaboration with the Stanford Palumbi Lab. We manually labelled a dataset of 1000 images and trained CNNs for coral bleaching classification. Our classifier is able to distinguish between bleached and unbleached coral with 98% accuracy, and between 5 different bleaching categories with 60% accuracy.
Here's a poster we presented on our work.
