News:

---

• [2024/10/18] The arxiv paper is available.
• [2024/10/13] Update the code and scripts.

1. 创建虚拟环境

---

安装必要的依赖torch=1.10.2,torchvision=0.11.3 和 mmcv-full=1.5.0

推荐使用conda虚拟环境 shell conda create -n SiamSeg python==3.8 -y conda activate SiamSeg pip 命令安装torch && torchvision && mmcv-full shell pip install torch==1.10.2+cu111 torchvision==0.11.3+cu111 -f https://download.pytorch.org/whl/torch_stable.html pip install mmcv-full==1.5.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.10.0/index.html pip install kornia matplotlib prettytable timm yapf==0.40.1 for CN user: shell pip install torch==1.10.2+cu111 -f https://mirror.sjtu.edu.cn/pytorch-wheels/cu111/?mirror_intel_list pip install torchvision==0.11.3+cu111 -f https://download.pytorch.org/whl/torch_stable.html pip install mmcv-full==1.5.0 -f https://download.openmmlab.com/mmcv/dist/cu111/torch1.10.0/index.html pip install kornia matplotlib prettytable timm yapf==0.40.1 ` 安装参考文档refer:

torch与torchvision版本对照

mmcv的版本对照

2.数据集的准备

---

我们选择 Postsdam, Vaihingen 和 LoveDA 作为基准数据集, 并创建了 train, val, test 列表供研究人员参考.

2.1 文件的下载

ISPRS Potsdam

The Potsdam dataset is for urban semantic segmentation used in the 2D Semantic Labeling Contest - Potsdam.

The dataset can be requested at the challenge homepage. The '2OrthoRGB.zip', '3OrthoIRRG.zip' and '5Labelsall_noBoundary.zip' are required.

ISPRS Vaihingen

The Vaihingen dataset is for urban semantic segmentation used in the 2D Semantic Labeling Contest - Vaihingen.

The dataset can be requested at the challenge homepage. The 'ISPRSsemanticlabelingVaihingen.zip' and 'ISPRSsemanticlabelingVaihingengroundtrutherodedCOMPLETE.zip' are required.

LoveDA

The data could be downloaded from Google Drive here.

Or it can be downloaded from zenodo, you should run the following command:

```shell

cd /{yourprojectbase_path}/SiamSeg/data/LoveDA

Download Train.zip

wget https://zenodo.org/record/5706578/files/Train.zip

Download Val.zip

wget https://zenodo.org/record/5706578/files/Val.zip

Download Test.zip

wget https://zenodo.org/record/5706578/files/Test.zip ```

2.2 数据集预处理

将下载的文件放入对应的路径中 格式如下 ```text SiamSeg/ ├── data/ │ ├── LoveDA/ │ │ ├── Test.zip │ │ ├── Train.zip │ │ └── Val.zip ├── ├── PotsdamIRRGDA/ │ │ ├── 3OrthoIRRG.zip │ │ └── 5LabelsallnoBoundary.zip ├── ├── PotsdamRGBDA/ │ │ ├── 2OrthoRGB.zip │ │ └── 5LabelsallnoBoundary.zip ├── ├── PotsdamIRRGDA/ │ │ ├── ISPRSsemanticlabelingVaihingen.zip │ │ └── ISPRSsemanticlabelingVaihingengroundtrutherodedCOMPLETE.zip

- Potsdam shell python tools/convertdatasets/potsdam.py data/PotsdamIRRG/ --clipsize 512 --stridesize 512 python tools/convertdatasets/potsdam.py data/PotsdamRGB/ --clipsize 512 --stridesize 512 - Vaihingen shell python tools/convertdatasets/vaihingen.py data/VaihingenIRRG/ --clipsize 512 --stridesize 256 - LoveDA shell cd data/LoveDA unzip Train.zip, Val.zip, Test.zip ```

3.Training

---

3.1 预训练模型的准备

mitb5.pth : We provide a script mit2mmseg.py in the tools directory to convert the key of models from the official repo to MMSegmentation style. ```shell python tools/modelconverters/mit2mmseg.py ${PRETRAIN_PATH} ./pretrained ``` 或者可以从 google drive 下载

文件的结构如下 text SiamSeg/ ├── pretrained/ │ ├── mit_b5.pth (needed) │ └── ohter.pth (option)

3.2 Potsdam IRRG to Vaihingen IRRG

tips

在linux下使用分布式训练脚本时, 由于文件权限的问题, 需要设置下训练脚本的权限 shell cd SiamSeg chmod 777 ./tools/dist_train.sh chmod 777 ./tools/dist_test.sh

```shell

Potsdam IRRG to Vaihingen IRRG

CUDAVISIBLEDEVICES 可见的GPU id 是 0-3 总计四个GPU处理器

PORT 设置分布式训练时master的通信端口

最后的4表示使用的GPU总数

CUDAVISIBLEDEVICES=0,1,2,3 PORT=10985 \ ./tools/disttrain.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamIRRG2_VaihingenIRRG.py 4 ```

3.3 Potsdam RGB to Vaihingen IRRG

```shell

Potsdam IRRG to Vaihingen IRRG

CUDAVISIBLEDEVICES 可见的GPU id 是 0-3 总计四个GPU处理器

PORT 设置分布式训练时master的通信端口

最后的4表示使用的GPU总数

CUDAVISIBLEDEVICES=0,1,2,3 PORT=10985 \ ./tools/disttrain.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamRGB2_VaihingenIRRG.py 4

```

3.4 Vaihingen IRRG to Potsdam IRRG

```shell

Potsdam IRRG to Vaihingen IRRG

CUDAVISIBLEDEVICES 可见的GPU id 是 0-3 总计四个GPU处理器

PORT 设置分布式训练时master的通信端口

最后的4表示使用的GPU总数

CUDAVISIBLEDEVICES=0,1,2,3 PORT=10985 \ ./tools/disttrain.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2_PotsdamIRRG.py 4 ```

3.5 Vaihingen IRRG to Potsdam RGB

```shell

Potsdam IRRG to Vaihingen IRRG

CUDAVISIBLEDEVICES 可见的GPU id 是 0-3 总计四个GPU处理器

PORT 设置分布式训练时master的通信端口

最后的4表示使用的GPU总数

CUDAVISIBLEDEVICES=0,1,2,3 PORT=10985 \ ./tools/disttrain.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2_PotsdamRGB.py 4 ```

3.6 LoveDA Rural to Urban

```shell

Potsdam IRRG to Vaihingen IRRG

CUDAVISIBLEDEVICES 可见的GPU id 是 0-3 总计四个GPU处理器

PORT 设置分布式训练时master的通信端口

最后的4表示使用的GPU总数

CUDAVISIBLEDEVICES=0,1,2,3 PORT=10985 \ ./tools/disttrain.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kRural2_Urban.py 4 ```

4.Testing

4.1 Potsdam IRRG to Vaihingen IRRG

```shell

for dist test

CUDAVISIBLEDEVICES=4,5,6,7 PORT=10985 \ sh tools/disttest.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamRGB2VaihingenIRRG.py \ {beastmodel_path} 4 --eval mIoU mFscore

for predict label save

launcher must set to 'none'

opacity between 0 and 1

PYTHONPATH=$(pwd):$PYTHONPATH CUDAVISIBLEDEVICES=0 \ python tools/test.py \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamRGB2VaihingenIRRG.py \ {beastmodelpath} --eval mIoU mFscore --launcher none --opacity 1.0 ```

4.2 Potsdam RGB to Vaihingen IRRG

```shell

for dist test

CUDAVISIBLEDEVICES=4,5,6,7 PORT=10985 \ sh tools/disttest.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamRGB2VaihingenIRRG.py \ {beastmodel_path} 4 --eval mIoU mFscore

for predict label save

launcher must set to 'none'

opacity between 0 and 1

PYTHONPATH=$(pwd):$PYTHONPATH CUDAVISIBLEDEVICES=0 \ python tools/test.py \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kPotsdamRGB2VaihingenIRRG.py \ {beastmodelpath} --eval mIoU mFscore --launcher none --opacity 1.0 ```

4.3 Vaihingen IRRG to Potsdam IRRG

```shell

for dist test

CUDAVISIBLEDEVICES=4,5,6,7 PORT=10985 \ sh tools/disttest.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2PotsdamIRRG.py \ {beastmodel_path} 4 --eval mIoU mFscore

for predict label save

launcher must set to 'none'

opacity between 0 and 1

PYTHONPATH=$(pwd):$PYTHONPATH CUDAVISIBLEDEVICES=0 \ python tools/test.py \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2PotsdamIRRG.py \ {beastmodelpath} --eval mIoU mFscore --launcher none --opacity 1.0 ```

4.4 Vaihingen IRRG to Potsdam RGB

```shell

for dist test

CUDAVISIBLEDEVICES=4,5,6,7 PORT=10985 \ sh tools/disttest.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2PotsdamRGB.py \ {beastmodel_path} 4 --eval mIoU mFscore

for predict label save

for predict label save

launcher must set to 'none'

opacity between 0 and 1

PYTHONPATH=$(pwd):$PYTHONPATH CUDAVISIBLEDEVICES=0 \ python tools/test.py \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kVaihingenIRRG2PotsdamRGB.py \ {beastmodelpath} --eval mIoU mFscore --launcher none --opacity 1.0 ```

4.5 LoveDA Rural to Urban

```shell

for dist test

CUDAVISIBLEDEVICES=4,5,6,7 PORT=10985 \ sh tools/disttest.sh \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kRural2Urban.py \ {beastmodel_path} 4 --eval mIoU mFscore

for predict label save

launcher must set to 'none'

opacity between 0 and 1

PYTHONPATH=$(pwd):$PYTHONPATH CUDAVISIBLEDEVICES=0 \ python tools/test.py \ configs/SiamSeg/siamsegdaformersepasppmitb5512x51240kRural2Urban.py \ {beastmodelpath} --eval mIoU mFscore --launcher none --opacity 1.0 ```

References

---

Many thanks to their excellent works * MMSegmentation * DACS * DAFormer