Perception Action Cycle-based Automatic Coral Classification for Underwater Survey Images

Principal Investigator: José Príncipe

Sponsor: Florida Atlantic University

Start Date: July 1, 2016

End Date: June 30, 2020

Amount: $156,000

Abstract

The purpose of this project is to develop automated ways to quantify the percentage of coral in reefs from long videos taken from underwater ROVs. We are using a semi-supervised machine learning technique to accomplish the task because the amount of labeled data is minimal (40 pixels per frame in one hundred of frames). If we are successful, we can automate the procedure to actually track the changes due to global warming.