Mastermine Mining-Site Object Detection Dataset (TERNA MAG – BOBAKAS Mine)

The dataset was developed as part of the EU-funded Mastermine project, designed for training and evaluating machine learning models for visual recognition of construction site elements and heavy machinery. The dataset was recorded at BOBAKAS mine in TERNA MAG facilities. It contains annotated images covering 8 object classes commonly found in mining and excavation environments: bulldoze car driller dump_truck excavator grader human truck The dataset includes 8 classes with bounding-box annotations following the format used by standard object-detection frameworks. Images are split into train, validation, and test sets, located in the following directories: train/images valid/images test/images The dataset split is: 594 train, 129 valid and 81 test images. Annotations follow the same folder structure under their respective labels/ directories. This dataset was originally prepared and hosted through Roboflow, under the project “mastermine”, and was downloaded with TXT annotations and a YAML config file used with YOLOv8. Roboflow project reference: Workspace: duth-41ltl Project: mastermine Version: 1 URL: https://universe.roboflow.com/duth-41ltl/mastermine/dataset/1 More datasets with augmentation and preprocessing are available on the Roboflow platform.