remote-sensing

UniMatch for Remote Sensing Change Detection

We provide the official PyTorch implementation of our UniMatch in the scenario of semi-supervised remote sensing change detection:

Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation
Lihe Yang, Lei Qi, Litong Feng, Wayne Zhang, Yinghuan Shi
In Conference on Computer Vision and Pattern Recognition (CVPR), 2023

Results

You can refer to our training logs for convenient comparisons during reproducing.

WHU-CD

The two numbers in each cell denote the changed-class IoU and overall accuracy, respectively.

Method	5%	10%	20%	40%
S4GAN	18.3 / 96.69	62.6 / 98.15	70.8 / 98.60	76.4 / 98.96
SemiCDNet	51.7 / 97.71	62.0 / 98.16	66.7 / 98.28	75.9 / 98.93
SemiCD	65.8 / 98.37	68.1 / 98.47	74.8 / 98.84	77.2 / 98.96
UniMatch (PSPNet)	77.5 / 99.06	78.9 / 99.10	82.9 / 99.26	84.4 / 99.32
UniMatch (DeepLabv3+)	80.2 / 99.15	81.7 / 99.22	81.7 / 99.18	85.1 / 99.35

LEVIR-CD

The two numbers in each cell denote the changed-class IoU and overall accuracy, respectively.

Method	5%	10%	20%	40%
S4GAN	64.0 / 97.89	67.0 / 98.11	73.4 / 98.51	75.4 / 98.62
SemiCDNet	67.6 / 98.17	71.5 / 98.42	74.3 / 98.58	75.5 / 98.63
SemiCD	72.5 / 98.47	75.5 / 98.63	76.2 / 98.68	77.2 / 98.72
UniMatch (PSPNet)	75.6 / 98.62	79.0 / 98.83	79.0 / 98.84	78.2 / 98.79
UniMatch (DeepLabv3+)	80.7 / 98.95	82.0 / 99.02	81.7 / 99.02	82.1 / 99.03

Getting Started

Installation

cd UniMatch
conda create -n unimatch python=3.10.4
conda activate unimatch
pip install -r requirements.txt
pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 -f https://download.pytorch.org/whl/torch_stable.html

Pretrained Backbone

• ResNet-50

├── ./pretrained
    └── resnet50.pth

Dataset

• WHU-CD: imageA, imageB, and label
• LEVIR-CD: imageA, imageB, and label

Please modify your dataset path in configuration files.

├── [Your WHU-CD/LEVIR-CD Path]
    ├── A
    ├── B
    └── label

Usage

UniMatch

# use torch.distributed.launch
# switch to current folder
sh scripts/train.sh <num_gpu> <port>

To train on other datasets or splits, please modify dataset and split in train.sh.

Supervised Baseline

Modify the method from 'unimatch' to 'supervised' in train.sh, and double the batch_size in configuration file if you use the same number of GPUs as semi-supervised setting (no need to change lr).

Citation

If you find this project useful, please consider citing:

@inproceedings{unimatch,
  title={Revisiting Weak-to-Strong Consistency in Semi-Supervised Semantic Segmentation},
  author={Yang, Lihe and Qi, Lei and Feng, Litong and Zhang, Wayne and Shi, Yinghuan},
  booktitle={CVPR},
  year={2023}
}

Acknowledgement

The processed WHU-CD and LEVIR-CD datasets are borrowed from SemiCD.