WIO-ReefFish: A High-Resolution Dataset for Taxon-Aware Coral Reef Fish Detection in the Western Indian Ocean

WIO-ReefFish is a high-resolution computer-vision dataset for coral reef fish detection in the Western Indian Ocean. The dataset was derived from diver-operated video (DOV) transects collected in Kenya in 2023 and Tanzania in 2025. Version 1.0.0 contains 1,000 ultra-high-definition images (3840 × 2160 pixels) and 6,768 bounding-box annotations across 24 taxonomic categories, corresponding to 23 fish families and one subfamily, Scarinae. The dataset is organised into a headline random train/validation/test split of 700/150/150 images and includes alternative benchmark splits for cross-transect and cross-country generalisation analyses. The repository includes YOLO-format image labels, class mappings, metadata, benchmark split files, and supporting documentation. WIO-ReefFish is intended to support object detection, class-aware reef fish detection, class-agnostic fish localisation, and benchmarking of computer-vision methods for coral reef biodiversity monitoring.