Jadhav Rohini, Thite Sandip, Pawar Shital, Patil Kailas, Chumchu Prawit
Bharati Vidyapeeth (Deemed to be University) College of Engineering, Pune, India.
Vishwakarma University, Pune, India.
Data Brief. 2024 Aug 24;57:110873. doi: 10.1016/j.dib.2024.110873. eCollection 2024 Dec.
The Natural Pothole Dataset within River Environments is an extensive collection of 3992 high-resolution images [1] documenting various natural potholes located in riverine settings. Each image has been rigorously annotated utilizing the YOLO (You Only Look Once) object detection framework, which ensures precise bounding box coordinates and accurate class labels for identified potholes. The annotations are provided in XML format, facilitating seamless integration with machine learning algorithms and computer vision applications. This dataset is particularly valuable for researchers and professionals in Geomorphology, Hydrology, River Science, Machine Learning, Environmental Science, and geospatial analysis, offering a robust foundation for tasks such as pothole detection, classification, and predictive modelling. By focusing exclusively on the natural occurrence of potholes, the dataset captures the diversity in shapes, sizes, and environmental contexts, thereby enriching the study and understanding of riverine geomorphological processes.
河流环境中的天然坑洼数据集是一个包含3992张高分辨率图像的广泛集合[1]，记录了位于河流环境中的各种天然坑洼。每张图像都使用YOLO（You Only Look Once）目标检测框架进行了严格注释，该框架可确保为识别出的坑洼提供精确的边界框坐标和准确的类别标签。注释以XML格式提供，便于与机器学习算法和计算机视觉应用程序无缝集成。该数据集对于地貌学、水文学、河流科学、机器学习、环境科学和地理空间分析领域的研究人员和专业人员特别有价值，为坑洼检测、分类和预测建模等任务提供了坚实的基础。通过专门关注坑洼的自然出现情况，该数据集捕捉了形状、大小和环境背景的多样性，从而丰富了对河流地貌过程的研究和理解。
