Pavementscapes
收藏资源简介:
Pavementscapes是一个大规模的分层图像数据集,由东南大学交通学院创建,用于沥青路面损伤分割。该数据集包含4000张分辨率为1024×2048的图像,这些图像来自中国15个不同路面的实际检测项目。数据集中的8680个损伤实例在像素级别上手动标记,涵盖6种损伤类别。Pavementscapes数据集旨在通过深度神经网络开发和评估路面损伤分割方法,解决现有公共数据集在路面损伤分割领域的局限性。
Pavementscapes is a large-scale hierarchical image dataset created by the School of Transportation, Southeast University, for asphalt pavement damage segmentation. This dataset contains 4,000 images with a resolution of 1024×2048, which are collected from actual pavement detection projects across 15 different road surfaces in China. A total of 8,680 damage instances in the dataset are manually annotated at the pixel level, covering 6 damage categories. The Pavementscapes dataset aims to develop and evaluate pavement damage segmentation methods via deep neural networks, addressing the limitations of existing public datasets in the field of pavement damage segmentation.




