MERMAID AI: Image Classification (Beta) Demo Webinarv

Alexandra Kler Lago, Field Support & Partnerships Manager, as we unveil the latest updates to MERMAID.

Through Mermaid AI do the following:

• Upload photo quadrats directly into MERMAID
• Centralize your data in one place
• Reduce time spent training models

We’re excited to share our newest beta features, developed in collaboration with CoralNet, that allow you to do just that.

Alex provides a live demonstration of the new features, including how you can now upload photos, automatically process them with MERMAID AI, and visualize your photo quadrat data alongside your other observations in our platform.

MERMAID is an initiative of the Wildlife Conservation Society (WCS). The WCS Global Conservation Program is the world's largest conservation field program, protecting more than 50 percent of Earth's known biodiversity, in partnership with governments, Indigenous People, local communities, and the private sector in more than 50 countries.

Webinar Summary:

On World Oceans Day 2025, MERMAID launched its new AI-powered image classification tool, now available in beta. This innovative feature allows users to analyze benthic photo quadrats directly within MERMAID Collect using a shared AI model developed in collaboration with CoralNet, the University of California San Diego, and Scripps Institution of Oceanography.

Trained on over 16 million publicly available images, the MERMAID AI model automates the classification of benthic organisms into 54 categories, including 12 benthic groups and 37 coral genera with growth forms. Users can drag and drop quadrat images into MERMAID, where AI suggests labels for each of 25 classification points per image. The system supports real-time human verification, enabling efficient and accurate data processing while improving the model through continued use.

The feature is designed to streamline workflows, enhance data consistency across projects, and accelerate reporting — with a vision to scale coral reef monitoring globally by 2035. The beta release is open for feedback from the community to help refine the tool ahead of full launch.