HSR-FP-SEG

We provide an open-access and well-annotated sample dataset for semantic segmentation of floatable plastics along Beijing-Shanghai high-speed railway of China, HSR-FP-SEG, which includes three categories of plastic mulching films, plastic greenhouses, plastic dust-proof nets.The VHR satellite imagery used is derived from Google Earth at 18-level with a spatial resolution of 1 m/pixel. Before labelling, all the VHR images are cropped into 512×512 patches. LabelImg was used for the pixel-wise annotation of all the floatable plastics. Meanwhile, the ratio of plastic mulching films, plastic greenhouses and plastic dust-proof nets is set to be 1:1:1, each with 500 samples and resulting in a total of 1,500 samples. Afterwards, the total sample dataset is split into training set, validation set and test set with a ratio of 7:1:2.

本研究公开了一款面向中国京沪高铁沿线漂浮塑料语义分割任务的开放获取且标注规范的样本数据集HSR-FP-SEG。该数据集包含塑料农用地膜、塑料大棚膜、塑料防尘网三类标注类别。所用极高分辨率（Very High Resolution, VHR）卫星影像取自谷歌地球（Google Earth）的18级影像，空间分辨率为1米每像素。标注前，所有VHR影像均被裁剪为512×512的图像块。本数据集采用LabelImg工具完成所有漂浮塑料的像素级标注。同时，三类样本的采样比例设置为1:1:1，每类各含500份样本，总计1500份样本。随后，全部数据集按照7:1:2的比例划分为训练集、验证集与测试集。

搜集汇总

背景与挑战

背景概述

HSR-FP-SEG是一个用于语义分割的开放访问数据集，包含三类可漂浮塑料（塑料地膜、塑料大棚和塑料防尘网）的标注，共1500个样本。数据集基于Google Earth的VHR卫星影像（1米/像素分辨率），并按7:1:2的比例分为训练集、验证集和测试集。