Mixed Medaka Ricefish (MMR) Dataset

Data and code accompanying the article Medaka-EL: A Robust Ensemble Learning Framework for the Detection of Indonesian Ricefish under Data Scarcity. The repository contains the bounding-box annotations of the Mixed Medaka Ricefish (MMR) dataset (1,139 images; 1,239 instances) for Oryzias celebensis (class 0) and Oryzias javanicus (class 1) in YOLO format, the group-aware stratified hold-out and 5-fold assignments, the trained YOLOv8n fold models, the per-image predictions, and the scripts for ensemble fusion (NMS, Soft-NMS, NMW and WBF), bootstrap significance testing, robustness and runtime evaluation. Web images whose licences do not permit redistribution are provided as source URLs, licence information and file hashes with a download script.