CNNs offer a layered architecture for image classification, consisting of feature detectors, pooling layers, and dense classifiers. These systems extract increasingly abstract features from input data, making them ideal for visual recognition tasks in natural environments like coral reefs.

This research addressed multiple machine learning concerns: overfitting, limited training data, preprocessing bias, and real-world applicability. Using over 10,000 images captured at La‘aloa Beach Park, I trained species-specific CNN models that achieved over 82% accuracy for Pocillopora and over 86% for Porites.

This work localizes the challenge to Hawaiian reef systems, contributing novel domain-specific insight to both marine biology and AI model transferability.

• State Science Fair Finalist — Excellence in environmental science and machine learning.
• International Science and Engineering Fair (ISEF) — Presented to a global audience.
• $5,000 in Awards:
  • NOAA “Taking the Pulse of the Planet” — For environmental monitoring innovation.
  • Purple Maiʻa STEM Innovation Award — For technology rooted in Hawaiian ecology.

Coral reefs are biologically vital ecosystems and deeply woven into the culture and identity of Hawai‘i. From providing habitat for marine life to protecting coastlines, they are irreplaceable.

Yet they face growing threats—warming seas, acidification, and pollution. By using technology to better understand and monitor these reefs, we can inform efforts to protect them.

This project was born out of my personal love for the ocean, my commitment to science, and a belief that data can be a powerful force for conservation.