Deep learning–assisted coral morphology cover across 742 sites on the Great Barrier Reef (2022–2025): A Great Reef Census citizen science dataset

Abstract Citizen science can expand the spatial extent of coral reef surveys, but its value is often limited by inconsistent data quality and slow manual analysis. Here we combine volunteer‐collected imagery with deep learning to overcome these constraints, delivering validated coral cover data at scale. We present 46,879 benthic seascape images from 742 sites across 267 reefs on the Great Barrier Reef, collected by Great Reef Census volunteers between September 2022 and February 2025. Images were analyzed using a deep learning segmentation model combined with online labelling by citizen scientists, yielding site‐level percent cover for three key coral types: branching Acropora , plating Acropora , and massive‐form corals. Data were validated with 99% accuracy against manual annotation, with 18–80 images per site yielding 95% accuracy. This geolocated dataset complements professional and government reef monitoring and is publicly available for research, conservation planning, and further tool development.