DEAC underwater camera reef, halo, and seagrass photos 2018-2019

This dataset contains over 85,000 underwater images collected between March 2018 and March 2019 at Lighthouse Reef Atoll, Belize. All images were captured by Cuddeback trail cameras (model 1231) in hand-built underwater housings. Cameras took images every 15 minutes during daylight hours over multiple weeks- and months-long deployments. Cameras were paired at patch reef sites throughout the atoll, one camera at the edge of the sand halo around the relevant patch reef, one camera placed twice the halo's width away in the surrounding seagrass or macroalgae. Approximately half of the cameras were deployed in the Half Moon Caye marine protected area (MPA), while the others were deployed elsewhere throughout the atoll. All images have been named according to their capture site, date, and time using the camtrapR package (Niedballa et al. 2016) in R (R Core Team 2020). For most sites, this name designates protected area status (MPA or OUT), surrounding vegetation (g for grass or a for algae), site number within those conditions, and camera location (H for halo edge or C for seagrass), e.g. MPAg2H. Aside from renaming, images received no additional editing or post-processing. Images are grouped in .zip archives by site and collection period. These file archives can be extracted using 7-Zip or any similar free utility. The SortedFish.zip archives contain over 9,000 images hand-classified by researchers as containing at least one fish (\"Yes\") or containing no fish (\"No\"). These are further organized into folders by protected area (IN or OUT) and location (H for halo, C for seagrass/macroalgae). SortedFish.txt is a text file with a list of these same image file names and their folder locations within the SortedFish archives. These sorted images are a subset of the larger dataset and therefore will appear in other folders as well. Image quality declined over time at many sites due to overgrowth by fouling organisms. Because of this, images deemed completely unsuitable for determining the presence/absence or identity of fish and other marine animals were removed from the dataset, although the images included here still vary in quality. Further details of image collection, processing, and purpose can be found in \"A low-cost, long-term underwater camera trap network coupled with deep residual learning image analysis\" by Bilodeau et al. 2022 in PLOS ONE https://doi.org/10.1371/journal.pone.0263377.