Neural Network Classification of Corals

A major challenge of contemporary ecology is the documentation of ecosystem change over time.
Coral reefs are the most diverse and complex of marine ecosystems. Furthermore, coral reefs are
undergoing a severe decline worldwide resulting from the adverse synergistic influences of global
climate change, ocean acidification, and seawater warming, exacerbated by anthropogenic
eutrophication and pollution. These factors affect the decline in live reef cover and the decrease in
coral species, taking place at different rates and grades of severity, depending on site exposure to
stressors. Any remediation measures require extensive monitoring at multiple sites, for long periods
and short intervals. That time-consuming, tedious, manual classification of coral species and their
abundance in real time, in several reefs, is a challenging nearly impossible task. Deep learning (DL)
has unique properties for streamlining the description, analysis and monitoring of coral reefs, saving
time, and obtaining higher reliability and accuracy compared with error-prone human performance,
and for handling and analyzing the vast amounts of resulting information Two main reef-health
indicators are live coral cover and coral biodiversity. The core aim of this review is to underscore the
strength and reliability of the deep learning approach for documenting coral reef features, based on
the evaluation of the published application of this method to the description of coral reefs and their
species assemblages. We review the current developments in the field, outline its current limitations
and future developments.