Natural Pothole Dataset within River Environments

The Natural Pothole Dataset comprises 3992 high-resolution images capturing various instances of natural potholes found in river water. Each image is meticulously annotated using the YOLO (You Only Look Once) object detection framework, providing precise bounding box coordinates and corresponding class labels for the detected potholes. Annotations are provided in XML format, enabling easy integration with machine learning algorithms and computer vision pipelines. The dataset focuses exclusively on natural potholes, emphasizing their diverse shapes, sizes, and environmental contexts. Researchers and practitioners in fields such as Geomorphology, Geomorphology, Hydrology, River Science, Machine Learning, Environmental Science, computer vision, and geospatial analysis can leverage this dataset for tasks including pothole detection, classification, and predictive modeling.