Publication Date
Fall 2025
Degree Type
Master's Project
Degree Name
Master of Science in Computer Science (MSCS)
Department
Computer Science
First Advisor
Philip Heller
Second Advisor
Amith Kamath Belman
Third Advisor
Maya DeVries
Keywords
Coral Segmentation, Mask R-CNN, Computer Vision, Deep Learning, Florida Aquarium, Coral Propagation, Coral Detection
Abstract
Coral reef ecosystems have been drastically declining due to environmental stressors like climate change, diseases, and pollution, creating a need for restoration efforts. In the United States of America, the Florida Aquarium has been making a great effort to restore coral reefs through its coral propagation programs. This project demonstrates the potential to make the Florida Aquarium restoration efforts easier and faster by using Mask R-CNN for consistent, scalable, and automated coral identification. Deep Learning Models, such as Mask R-CNN and other segmentation frameworks, are effective tools for coral recognition as well as recording coral’s morphological changes. In this project, using the Coral Vision platform, a Mask R-CNN model was trained to detect coral fragments in images that came from a coral nursery at the Florida Aquarium. Out of 360 provided coral images, 30 were manually masked for training. The model showed high detection accuracy for the majority of the images, with confidence scores ranging between 97% to 100%. Some images, however, showed segmentation inconsistencies, indicating room for model refinement.
Recommended Citation
Sadre Arhami, Sadaf, "AUTOMATED CORAL DETECTION USING MASK R-CNN FOR RESTORATION MONITORING AT THE FLORIDA AQUARIUM" (2025). Master's Projects. 1592.
DOI: https://doi.org/10.31979/etd.gzut-c78k
https://scholarworks.sjsu.edu/etd_projects/1592
