A new urban village dataset in our recent work "Fine-Grained Urban Village Extraction By Mask Transformer From High-Resolution Satellite Images in Pearl River Delta" on IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.
The dataset contains 2102 pairs of high-resolution images of size 512×512 from Pearl River Delta in 2020 .The process of constructing the urban village dataset can be summarized as follows:
-
16 training images and 10 testing images are evenly and randomly selected from the 329 Google Earth images in Pearl River Delta
-
The 26 selected images are annotated through manual visual interpretation to obtain UV.
-
All vectors of UV annotated in the above steps will be rasterized into pixel-wise annotations, where the non-urban village area and the urban village area are denoted by 0 and 255, respectively.
MaskUV
- BaiduYun (qja6)
If you find our work useful for your research, please consider citing our paper:
@ARTICLE{10623471,
author={Chai, Zhuoqun and Liu, Mengxi and Shi, Qian and Zhang, Yuanyuan and Zuo, Minglin and He, Da},
journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
title={Fine-Grained Urban Village Extraction by Mask Transformer From High-Resolution Satellite Images in Pearl River Delta},
year={2024},
volume={17},
number={},
pages={13657-13668},
keywords={Feature extraction;Urban areas;Transformers;Training;Buildings;Rivers;Remote sensing;Deep learning;Pearl River Delta (PRD);remote sensing;urban villages (UVs);urbanization},
doi={10.1109/JSTARS.2024.3434487}}