This repository is the code implementation of the paper BiFA: Remote Sensing Image Change Detection with Bitemporal Feature Alignment
The current branch has been tested on Linux system, PyTorch 1.12.0 and CUDA 12.1, supports Python 3.7.
If you find this project helpful, please give us a star ⭐️, your support is our greatest motivation.
🌟 2024.06.20 Revised the BiFA project.
- FC-EF (ICIP'2018)
- FC-Siam-diff (ICIP'2018)
- FC-Siam-conc (ICIP'2018)
- IFN (ISPRS'2020)
- SNUNet (GRSL'2021)
- SwinUnet (TGRS'2022)
- BIT (TGRS'2022)
- ChangeFormer (IGARSS'22)
- MSCANet (JSTARS'2022)
- Paformer (GRSL'2022)
- DARNet (TGRS'2022)
- ACABFNet (JSTARS'2023)
- BiFA (TGRS'2024)
- ......
- Updated more change detection methods
- Introduction
- Benchmark
- TODO
- Table of Contents
- Installation
- Dataset Preparation
- Model Training and Testing
- Citation
- License
- Contact Us
- Linux system, Windows, depending on
MMCVcan be installed - Python 3.6+, recommended 3.7
- PyTorch 1.10+ or higher, recommended 1.12.0
- CUDA 11.7 or higher, recommended 12.1
It is recommended to use Miniconda for installation. The following commands will create a virtual environment named bifa and install PyTorch. In the following installation steps, the default installed CUDA version is 12.1. If your CUDA version is not 12.1, please modify it according to the actual situation.
Note: If you are experienced with PyTorch and have already installed it, you can skip to the next section. Otherwise, you can follow the steps below.
Details
Step 0: Install Miniconda.
Step 1: Create a virtual environment named bifa and activate it.
conda create -n bifa python=3.7
conda activate bifaStep 2: Install dependencies.
pip install -r requirements.txtNote: Use the following command to install mmcv
pip install mmcv-full==1.6.2 -f https://download.openmmlab.com/mmcv/dist/cu113/torch1.12/index.htmlYou can download or clone the BiFA repository.
git clone git@github.com:zmoka-zht/BiFA.git
cd BiFAWe provide the method of preparing the remote sensing change detection dataset used in the paper.
- Data download link: WHU-CD Dataset PanBaiDu. Code:t2sb
- Data download link: LEVIR-CD Dataset PanBaiDu. Code:5ss2
- Data download link: LEVIR+-CD Dataset PanBaiDu. Code: xtj8
- Data download link: SYSU-CD Dataset PanBaiDu. Code: gk0w
- Data download link: CLCD-CD Dataset PanBaiDu. Code: bcml
- Data download link: DSIFN-CD Dataset PanBaiDu. Code: u8e9
You can also choose other sources to download the data, but you need to organize the dataset in the following format：
${DATASET_ROOT} # Dataset root directory, for example: /home/username/data/LEVIR-CD
├── A
│ ├── train_1_1.png
│ ├── train_1_2.png
│ ├──...
│ ├── val_1_1.png
│ ├── val_1_2.png
│ ├──...
│ ├── test_1_1.png
│ ├── test_1_2.png
│ └── ...
├── B
│ ├── train_1_1.png
│ ├── train_1_2.png
│ ├──...
│ ├── val_1_1.png
│ ├── val_1_2.png
│ ├──...
│ ├── test_1_1.png
│ ├── test_1_2.png
│ └── ...
├── label
│ ├── train_1_1.png
│ ├── train_1_2.png
│ ├──...
│ ├── val_1_1.png
│ ├── val_1_2.png
│ ├──...
│ ├── test_1_1.png
│ ├── test_1_2.png
│ └── ...
├── list
│ ├── train.txt
│ ├── val.txt
│ └── test.txt
All configuration for model training and testing are stored in the local folder config
python train.py --config/levir.json python test.py --config/levir_test_bifa.json - The model weights of BiFA are provided in the
experiments/pretrain - Segformer weight download link: Segformer PanBaiDu. Code:i81p
If you use the code or performance benchmarks of this project in your research, please refer to the following bibtex citation of BiFA.
@ARTICLE{10471555,
author={Zhang, Haotian and Chen, Hao and Zhou, Chenyao and Chen, Keyan and Liu, Chenyang and Zou, Zhengxia and Shi, Zhenwei},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={BiFA: Remote Sensing Image Change Detection With Bitemporal Feature Alignment},
year={2024},
volume={62},
number={},
pages={1-17},
keywords={Feature extraction;Task analysis;Remote sensing;Transformers;Interference;Decoding;Optical flow;Bitemporal interaction (BI);change detection (CD);feature alignment;flow field;high-resolution optical remote sensing image;implicit neural representation},
doi={10.1109/TGRS.2024.3376673}}
@ARTICLE{11268372,
author={Zhang, Haotian and Guo, Han and Chen, Keyan and Chen, Hao and Zou, Zhengxia and Shi, Zhenwei},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={FoBa: A Foreground-Background co-Guided Method and New Benchmark for Remote Sensing Semantic Change Detection},
year={2025},
volume={},
number={},
pages={1-1},
keywords={Semantics;Remote sensing;Transformers;Feature extraction;Annotations;Roads;Multitasking;Spatial resolution;Landsat;Land surface;Semantic change detection (SCD);foreground-background co-guided;bi-temporal interaction;mamba;new benchmark},
doi={10.1109/TGRS.2025.3636947}}
@ARTICLE{10902569,
author={Zhang, Haotian and Chen, Keyan and Liu, Chenyang and Chen, Hao and Zou, Zhengxia and Shi, Zhenwei},
journal={IEEE Transactions on Geoscience and Remote Sensing},
title={CDMamba: Incorporating Local Clues Into Mamba for Remote Sensing Image Binary Change Detection},
year={2025},
volume={63},
number={},
pages={1-16},
keywords={Feature extraction;Transformers;Remote sensing;Convolutional neural networks;Visualization;Artificial intelligence;Spatiotemporal phenomena;Computational modeling;Attention mechanisms;Computer vision;Bi-temporal interaction;change detection (CD);high-resolution optical remote sensing image;Mamba;state-space model},
doi={10.1109/TGRS.2025.3545012}}
This project is licensed under the Apache 2.0 License.
If you have any other questions❓, please contact us in time 👬