EvLab-SS

EVLab-SS dataset是由胡祥云教授提出并领导，由张米博士和其他EVLab成员创建的语义分割基准。该程序从年中2016年开始，2018年结束，它旨在为遥感社区的真实工程场景找到一个良好的体系结构。该数据集最初是按照《地理条件调查》 (编号: GDPJ 01-2013) 中的第一类标准进行注释的。它包括48,622，13539分别用于训练和验证的补丁。该数据集包含11个主要类别，即背景，农田，花园，林地，草原，建筑，道路，结构，挖掘桩，沙漠和水域 (标签0-10)。空间分辨率范围为每像素0.1m至2m。

The EVLab-SS dataset is a semantic segmentation benchmark proposed and led by Professor Hu Xiangyun, and created by Dr. Zhang Mi and other members of EVLab. The development of this dataset began in mid-2016 and concluded in 2018, aiming to explore high-performance architectures for real-world engineering scenarios in the remote sensing community. Initially, the dataset was annotated in accordance with the first-category standards stipulated in *Geographical Condition Survey* (No. GDPJ 01-2013). It includes 48,622 patches for training and 13,539 patches for validation. The dataset contains 11 major categories: background, farmland, garden, woodland, grassland, building, road, structure, excavation pile, desert, and water body (labels 0-10). Its spatial resolution ranges from 0.1 m to 2 m per pixel.

搜集汇总

背景与挑战

背景概述

EvLab-SS是一个由武汉大学胡祥云教授团队创建的遥感语义分割基准数据集，发布于2022年，旨在支持真实工程场景的架构研究。该数据集包含48,622个训练补丁和13,539个验证补丁，总计约18.9GB，涵盖11个主要类别（如农田、建筑、水域等），空间分辨率范围为每像素0.1米至2米，适用于高精度遥感图像分析。