2023.12.26: 似乎原作者已经上传了自己的代码,可以点击此链接查看
2023.12.26: It seems that the original authors have uploaded their own code, click on this link to view.
DSen2_CR神经网络的pytorch版本
Pytorch version of the DSen2_CR neural network.
- 官方下载地址 / Official Download Link: https://dataserv.ub.tum.de/index.php/s/m1554803
- 百度网盘 / Baidu Netdisk:
- 链接 / Link: https://pan.baidu.com/s/11lypfXe24byyk5FM8yZ1ZA
- 提取码 / Extraction Code:
fxps - 请注意,数据集很大,需要将所有压缩分卷下下来才能解压,建议预留500G磁盘空间,下载单个压缩包下来无法解压,如有任何问题,请在我的仓库下面提ISSUE,或者邮件联系我
请按照以下方式组织数据:
Please organize the data as follows:
dataset_name
├── ROIs1158_spring
│ ├── s1_*
│ ├── s2_*
│ └── s2_cloudy_*
├── ROIs1868_summer
│ ├── s1_*
│ ├── s2_*
│ └── s2_cloudy_*
├── ROIs1970_fall
│ ├── s1_*
│ ├── s2_*
│ └── s2_cloudy_*
└── ROIs2017_winter
├── s1_*
├── s2_*
└── s2_cloudy_*- 可以使用Visdom进行可视化。
You can use Visdom for visualization. - 打开CMD,输入以下命令以下载Visdom:
Open CMD and enter the following command to install Visdom:pip install visdom
- 启动Visdom服务器:
Start the Visdom server:
python -m visdom.server
- 在浏览器中查看训练图片和损失。 View training images and losses in your browser.
- 如果不想要这个功能,请在train.py中删除包含"vis"的代码。 If you don’t want this feature, delete the code containing "vis" in train.py.
- 在config.py中可以调整网络设置。
You can adjust the network settings in
config.py.
- 使用以下命令训练网络:
Use the following command to train the network
python train.py
- 使用以下命令进行预测:
Use the following command for prediction:
python predict.py
- 确保数据集已正确组织并放置在dataset_name目录下。 Ensure the dataset is organized correctly and placed in the dataset_name directory.
- 训练和预测过程中,请根据实际情况调整参数。 Adjust parameters as needed during training and prediction.