Submeter-level global-scale road mapping based on limited labeled data from optical remote sensing images

by Ning Zhou, Mingting Zhou, Xuanhao Wang, Dingyuan Chen, Weiyue Shi, Yihao Zhu, Junyi Liu and Haigang Sui

This is an official implementation of GlobalRoadMapper.

Environments:

• pytorch >= 1.11.0
• python >=3.6
• You can install the Python packages involved in the code yourself.

Dataset

your_source_dataset
    --train
        --image
        --label
    --test
        --image
        --label

your_target_dataset
    --train
        --image
    --test
        --image
        --label

Train

python train.py --config_path configs/config_train.json

Inference

python test.py --config_path configs/config_test_onlydir.json

Citation

If you use GlobalRoadMapper in your research, please cite the following paper.

@ARTICLE{Zhou2025GlobalRoadMapper,
  author={Zhou, Ning and Zhou, Mingting and Wang, Xuanhao and Chen, Dingyuan and Shi, Weiyue and Zhu, Yihao and Liu, Junyi and Sui, Haigang},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Submeter-Level Global-Scale Road Extraction Based on Limited Labeled Data From Optical Remote Sensing Images}, 
  year={2025},
  volume={63},
  number={4411323},
  pages={1-23},
  doi={10.1109/TGRS.2025.3572373}}