BAN

Official implementation for the paper "A New Learning Paradigm for Foundation Model-based Remote Sensing Change Detection", the code is developed on top of Open-CD v1.1.0.

News

• 2/10/2024 - BAN is supported in Open-CD. 😋

Usage

Install

• Create a conda virtual environment and activate it:

conda create -n BAN python=3.8 -y
conda activate BAN

• Install PyTorch and torchvision :

For examples, to install torch==2.0.0 with CUDA==11.8:

pip install torch==2.0.0 torchvision==0.15.1 torchaudio==2.0.1 --index-url https://download.pytorch.org/whl/cu118

• Install OpenMMLab Toolkits as Python packages:

pip install -U openmim
mim install mmengine==0.10.1
mim install mmcv==2.1.0
mim install mmpretrain==1.1.1
pip install mmsegmentation==1.2.2
pip install mmdet==3.2.0

• Install Open-CD

git clone https://github.com/likyoo/open-cd.git
cd open-cd
pip install -v -e .
cd ..

• Install other requirements:

pip install ftfy regex

• Clone this repo:

git clone https://github.com/likyoo/BAN.git
cd BAN

Data Preparation

Download datasets (LEVIR-CD, S2Looking, BANDON, WHU-CD) and move (or link) them to BAN/data.

Evaluation

To evaluate our BAN on LEVIR-CD test, run:

python test.py <config-file> <checkpoint>

You can download checkpoint files from huggingface | baidu disk.

For example, to evaluate the BAN-vit-l14-clip_mit-b0 with a single GPU:

python test.py configs/ban/ban_vit-l14-clip_mit-b0_512x512_40k_levircd.py checkpoint/ban_vit-l14-clip_mit-b0_512x512_40k_levircd.pth

Training

To train the BAN, run:

python train.py <config-file>

For example, to train the BAN-vit-l14-clip_mit-b0 with a single GPU on LEVIR-CD, run:

python train.py configs/ban/ban_vit-l14-clip_mit-b0_512x512_40k_levircd.py

Note: You can download pretrained files from huggingface | baidu disk.

Citation

@ARTICLE{10438490,
  author={Li, Kaiyu and Cao, Xiangyong and Meng, Deyu},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={A New Learning Paradigm for Foundation Model-based Remote Sensing Change Detection}, 
  year={2024},
  volume={},
  number={},
  pages={1-1},
  keywords={Adaptation models;Task analysis;Data models;Computational modeling;Feature extraction;Transformers;Tuning;Change detection;foundation model;visual tuning;remote sensing image processing;deep learning},
  doi={10.1109/TGRS.2024.3365825}}