EGMS-Net

Pytorch codes of Multi-task Siamese Network Guided by Enhanced Change Information for Semantic Change Detection in Bi-Temporal Remote Sensing Images [paper]

Dataset Download

In the following, we summarize the processed semantic change detection data set used in this paper:

• SECOND (Baidu)
• Landsat-SCD (Baidu)
• HiUCD-mini (Baidu)

How to Use

1. Dataset preparation.
   • Please split the data into training, validation and test sets and organize them as follows:

      YOUR_DATA_DIR
      ├── ...
      ├── train
      │   ├── A
      │   ├── B
      │   ├── labelA
      │   ├── labelB
      ├── val
      │   ├── A
      │   ├── B
      │   ├── labelA
      │   ├── labelB
      ├── test
      │   ├── A
      │   ├── B
      │   ├── labelA
      │   ├── labelB

2. Change the root path of the data set.

   • Take the SECOND data set as an example.
   • Find line 34 in EGMS-Net/train_SECOND.py, change --data_root to your local dataset directory.
   • Find line 34 and line 35 in EGMS-Net/inference_SECOND.py, change --data_root to your local dataset directory, and change --load_from to your checkpoint path.

3. Training

   bash my_run.sh

4. Inference and evaluation

   bash my_inference.sh

Pretrained Models

The reproducible weights of EGMS-Net on the three benchmark datasets are visible in the links below:

Baidu

Cite EGMS-Net

If you find this work useful or interesting, please consider citing the following BibTeX entry.

@ARTICLE{10737132,
  author={Zuo, Xibing and Jin, Fei and Ding, Lei and Wang, Shuxiang and Lin, Yuzhun and Liu, Bing and Ding, Yao},
  journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing}, 
  title={Multitask Siamese Network Guided by Enhanced Change Information for Semantic Change Detection in Bitemporal Remote Sensing Images}, 
  year={2025},
  volume={18},
  number={},
  pages={61-77},
  keywords={Feature extraction;Semantics;Multitasking;Training;Remote sensing;Periodic structures;Vectors;Head;Data mining;Attention mechanisms;Change information enhancement;change information guidance;multitask learning;remote sensing;semantic change detection (SCD)},
  doi={10.1109/JSTARS.2024.3487137}}