Fish-gres YOLOv8 Object Detection Annotations (Derived Release) v1.0

This record provides a detection-ready derivative release of the Fish-gres image dataset, containing manual bounding-box annotations in YOLOv8 format. Source images:Fish-gres Dataset for Fish Species Classification (Mendeley Data), DOI: 10.17632/76cr3wfhff.1, licensed under CC BY 4.0. What was added/changed:• Manual bounding-box annotation for single-fish object detection (one fish per image), created using Roboflow and exported in YOLOv8 label format.• Train/validation/test split files and a YOLO dataset configuration file (e.g., data.yaml). Dataset size and split:Total 2,000 images:• Train: 1,400 (70%)• Validation: 400 (20%)• Test: 200 (10%) Filename note:Because the dataset was exported via Roboflow, image filenames may differ from the original Fish-gres filenames. Labels match images by basename within each split folder (e.g., xxx.jpg ↔ xxx.txt). Attribution:Required citation: Fish-gres (images) and this Zenodo record (YOLOv8 annotations/splits).Recommended citation: our paper (IOP EES, 10.1088/1755-1315/1359/1/012023) if you rely on the experimental setup or wish to reference the study that used these annotations. This dataset release was uploaded after the article’s publication to support replication and reuse of the object-detection experiments reported in the paper.