EVLab-SS dataset is a semantic segmentation benchmark proposed and leaded by Prof. Xiangyun Hu and created by Dr. Mi Zhang and other EVLab members.
The program is started from the mid year of 2016 and end in 2018, which aims to find a good architecture for real engineered scenes in remote sensing community.
The dataset is originally annotated following the standard of the first class under the Geographic Conditions Survey (NO.GDPJ 01—2013). It comprises 48,622, 13539 patches for training and validation, respectively. The dataset contains 11 major classes, namely, background, farmland, garden, woodland, grassland, building, road, structures, digging pile, desert and waters (label 0-10). The spatial resolution ranges from 0.1m to 2m per pixel.
#######################################################################################
background - 0
farmland - 1
garden - 2
woodland - 3
grassland - 4
building - 5
road - 6
structures - 7
digging pile - 8
desert - 9
waters - 10
#######################################################################################
Download:
BaiduCloud：https://pan.baidu.com/s/1V7qdgoG0PpZrmsCzT3ROkQ
Code：EVLa
#######################################################################################
!!!!!! Notes: the benchmark is reserved for research purpose only, please do not use for other purposes,
including extending it to your own benchmark without our permission !!!!!!
Contact Information:
{huxy, mizhang}@whu.edu.cn
#######################################################################################
If you use the dataset in your research, please consider citing:
@article{Zhang2017Learning,
title={Learning Dual Multi-Scale Manifold Ranking for Semantic Segmentation of High-Resolution Images},
author={Zhang, Mi and Hu, Xiangyun and Zhao, Like and Lv, Ye and Luo, Min and Pang, Shiyan},
year={2017},
}
All copyrights reserved by @EarthVision Laboratory, School of Remote Sensing and Information Engineering, Wuhan University.