Global road extraction using a pseudo-label guided framework: From benchmark dataset to cross-region semi-supervised learning

Xiaoyan Lu, Yanfei Zhong, Zhuo Zheng, Junjue Wang, Dingyuan Chen, Yu Su and Liangpei Zhang

[Paper]

Semi-supervised road extraction framework

The GRSet and the global-scale validation set

Highlights

1. The GRSet contains 47210 samples from 121 cities across six continents: Europe, Africa, Asia, South America, Oceania, and North America. The area is 49503.27 km2.
2. The global-scale validation set covers more than 30 cities, across four continents: Europe, Africa, Asia, and North America. The area is 2836.38 km2.

Downloads

The GRSet and the global-scale validation set are released at Baidu Drive Code:

Pre-trained model

The GRSet pre-trained LinkNet50 is released at Baidu Drive Code:

Feel free to use these datasets and design your own models, and we are looking forward to your exciting results!

Shared resources

Massachusetts Road Dataset (2013)
RoadTracer Dataset (2018)
SpaceNet Road Dataset (2018)
DeepGlobe Road Dataset (2018)
LSRV: The Large-Scale Road Validation Dataset (2021)

If you are having difficulty processing this data, you can contact me at [email protected]

License

The owners of the data and of the copyright on the data are RSIDEA, Wuhan University. Use of the Google Earth images must respect the "Google Earth" terms of use. All images and their associated annotations in GRSet and the global-scale validation set can be used for academic purposes only, but any commercial use is prohibited.