STCNet

code for learning spatial-temporal consistency for satellite image sequence prediction

Summary

We provide the pretrained model and test script for 4-hour satellite nowcasting over the China mainland.

Comparsion Samples

The ground-truth 4-hour sequences:

Pred 4-hour satellite sequences:

Test

make satellite nowcasting over the China mainland area.

bash test_bash.sh

Citation

If you are interested in our repository or our paper, please cite the following papers:

@article{dai2023learning,
  title={Learning Spatial-Temporal Consistency for Satellite Image Sequence Prediction},
  author={Dai, Kuai and Li, Xutao and Ma, Chi and Lu, Shenyuan and Ye, Yunming and Xian, Di and Tian, Lin and Qin, Danyu},
  journal={IEEE Transactions on Geoscience and Remote Sensing},
  volume={61},
  pages={3303947},
  year={2023}
}