Underwater Crown-of-Thorn Starfish (COTS) imagery sample dataset

Full description

A small set of images collected from the Great Barrier Reef that can be used as a test/demo sample set for COTS detectors. It is a part of the complete dataset used for the Kaggle COTS detection competition (https://www.kaggle.com/c/tensorflow-great-barrier-reef). To use this data in scientific research, please cite the CSIRO COTS Dataset paper (https://arxiv.org/abs/2111.14311.).
Lineage: We have collected the data with GoPro Hero9 cameras that were adapted for use in the Manta Tow method. Specifically, the camera was attached to the bottom of the manta tow board that the snorkeler-diver holds onto during the surveys. The camera provides an oblique field of view of the reef below the diver and the distance to the reef dynamically changes as the diver explores the
reef. The distance is typically several meters from the bottom, but can be as close as a few tens of centimeters or as far as 10 meters or more. The survey boat moves at a speed up to 5 knots and it pulls the diver for a duration of two minutes which equates to transect of approximately 200 meters. The boat then stops to let the diver record data observed during the transect on a sheet of paper. We set the GoPro cameras to record videos continuously at 24 frames per second at 3840x2160 resolution and manually removed the periods
of no activity between transects.