Data and Mask-Evaluation Materials for Image-Based Estimation of Cubic Artificial Reef Settlement and Projected Remaining-Area Ratio Using Mask R-CNN

<p dir="ltr">This dataset supports the manuscript entitled “Image-Based Estimation of Cubic Artificial Reef Settlement and Projected Remaining-Area Ratio Using Mask R-CNN.” It includes the numerical data supporting the reported results, 14 unaugmented held-out test images with 38 annotated instances, the retained Mask R-CNN checkpoint, the mask-evaluation code and software requirements, and the resulting COCO-style mask-level evaluation metrics.

The held-out images were reported by the dataset curator to originate from more than ten source videos that were not used to prepare the development dataset. The exact historical mapping between individual test images and their source videos and the original internal training–validation subset lists were not retained. The included third-party Matterport Mask R-CNN source remains subject to its accompanying MIT License.</p>