A key component of addressing coral bleaching is accurate, cost-effective monitoring on appropriate spatial scales. This can be particularly challenging in remote places where access by experienced researchers is time consuming and expensive. Here, we tested whether a pre-program drone, RGB imagery and a deep learning algorithm could be used to accurately monitor coral bleaching on Lord Howe Island. Imagery was collected once by researchers and four times by a local pilot using a pre-programed flight plan across one year. Object-based image analysis was used to develop a training dataset, and a multiRes-Unet architecture was chosen for automated segmentation. A further analysis was conducted using segmented bleached coral to derive object rather than pixel classification, to improve accuracies and provide a count of individual bleached coral colonies. We determined that healthy corals can be classified with a high level of accuracy (Precision; 0.95, Recall; 0.83), and OBIA analysis of bleached corals achieved prediction accuracies of 0.72 in coral-dominant regions. Recovery-based classifications including algae covered coral and newly dead coral were not possible to assess accurately. Overall, RGB imagery collected using a pre-programed drone and local pilot combined with our deep learning algorithm can accurately monitor coral bleaching over entire reefs.