Assessment of Submerged Aquatic Vegetation in Lake Michigan Using Down-Looking AUV-Collected Imagery

This dataset comprises in situ images of benthic algae collected by an autonomous underwater vehicle (AUV) in Lake Michigan during August and September 2020. The images were annotated, or labeled, to identify specific features for use in machine learning and algorithm development, specifically to train and validate an algorithm for the automated identification and assessment of submerged aquatic vegetation (SAV). The dataset is organized into individual zip files, each corresponding to different collection dates, with the date indicated in the file name. Each zip file contains original images, ground truth binary masks where benthic algae are represented by white pixels and the background by black pixels, and JSON files specifying the coordinates of polygons delineating the algae in each image. Additionally, the metadata associated with these images is accessible in the accompanying metadata spreadsheet. The AUV, a modified L3Harris-OceanServer Iver3, was equipped with a high-performance 9-megapixel CMOS sensor, specifically the Allied Vision Manta G-895C color camera. Advanced sensors and imaging systems onboard the AUV enabled precise navigation and image capture during the survey. In total, over 300 km of lakebed were surveyed by the AUV across thirty sites along the Lake Michigan shoreline. The 4096 x 2176-pixel color images were georeferenced using the AUV's inertial navigation system (INS), which relied on an iXBlue PHINS Compact C3 fiber-optic gyroscope for accurate positioning. Each image was accompanied by recorded parameters, including temperature, altitude, depth, and time. Collectively, these data provide a comprehensive dataset for analyzing and understanding the distribution and dynamics of submerged aquatic vegetation (SAV) in Lake Michigan.