KyanChen/BuildingExtraction

STTNet

Paper: Building Extraction from Remote Sensing Images with Sparse Token Transformers

1. Prepare Data
   Prepare data for training, validation, and test phase. All images are with the resolution of $512 \times 512$. Please refer to the directory of Data.

   For larger images, you can patch the images with labels using Tools/CutImgSegWithLabel.py.

2. Get Data List
   Please refer to Tools/GetTrainValTestCSV.py to get the train, val, and test csv files.

3. Get Imgs Infos
   Please refer to Tools/GetImgMeanStd.py to get the mean value and standard deviation of the all image pixels in training set.

4. Modify Model Infos
   Please modify the model information if you want, or keep the default configuration.

5. Run to Train
   Train the model in Main.py.

6. [Optional] Run to Test
   Test the model with checkpoint in Test.py.

We have provided pretrained models on INRIA and WHU Datasets. The pt models are in folder Pretrain.

If you have any questions, please refer to our paper or contact with us by email.

@Article{rs13214441,
AUTHOR = {Chen, Keyan and Zou, Zhengxia and Shi, Zhenwei},
TITLE = {Building Extraction from Remote Sensing Images with Sparse Token Transformers},
JOURNAL = {Remote Sensing},
VOLUME = {13},
YEAR = {2021},
NUMBER = {21},
ARTICLE-NUMBER = {4441},
URL = {https://www.mdpi.com/2072-4292/13/21/4441},
ISSN = {2072-4292},
DOI = {10.3390/rs13214441}
}