Riley Frisk & Dr. Geoff Dilly
Abstract
The rocky intertidal zones on Santa Rosa Island are filled with diverse algae and invertebrates. Over the last seven years, our lab has used multiple methods to monitor that diversity. One of those techniques is photo-transect monitoring which utilizes AI machine learning for automated visual annotations of biological categories (e.g. filamentous green algae, membranous red/brown algae). In this experiment, we compared intertidal photos taken at two different sites on Santa Rosa Island, Bechers Bay and Skunk Point, across the years 2016, 2019, and 2022. Photos were taken using a Canon EOS camera mounted on a PVC rig and carried along 20m long transect lines running perpendicular to the ocean.
After collecting the images, we used the website CoralNet for all manual and automatic annotations as well as published established categorical labels using the CATAMI (Collaborative and Automated Tools for Analysis of Marine Imagery) classification scheme. The automated annotator was first compared to manual photo annotation across multiple transect lines. We modified the already trained classifier created by the Marine Biodiversity Observation Network (MBON) publicly available on CoralNet to fit our local Channel Islands species. After ensuring accuracy by comparing observer and AI biological category calls, we used the classifier to automatically annotate transects at both sites across all three years. The resulting data was run through an ANOVA analysis of biological categorical abundance over time, comparing the species changes between the two sites. Combining photo-transects with automated visual annotation is a fast and effective way to analyze the changing characteristics of our rocky intertidal sites on Santa Rosa Island.
Details
Session 1
11:15am – 12:30pm
Del Norte Hall
Room B: 1545
Leave a Reply