This is the official pytorch implementation for paper: Information-Bottleneck Driven Binary Neural Network for Change Detection, published in ICCV 2025.

• CUDA (version 11.6)
• Python3 (version 3.8)
• PyTorch

• PCD
• VL_CMU_CD
• LEVIR-CD

Please follow this site to download the PCD dataset. You may need to send e-mails to Prof. Ken Sakurada or Prof. Takayuki Okatani.

For VL_CMU_CD, please check this issue.

Please follow this site to download the LEVIR-CD dataset.

• PCD: googleDrive
• VL_CMU_CD: googleDrive
• LEVIR-CD: googleDrive

Our code refers to C-3PO and AERIS.

We use the training&evaluation framework for LEVIR-CD dataset from DGMA2-Net.

If you find the work useful for your research, please cite:

@inproceedings{yin2025bicd,
  title={Information-Bottleneck Driven Binary Neural Network for Change Detection},
  author={Yin, Kaijie and Zhang, Zhiyuan and Kong, Shu and Gao, Tian and Xu, Chengzhong and Kong, Hui},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  year={2025}
}

Any problem, please contact the first author (Email: yin.kaijie at connect.um.edu.mo)