TODO List • Version Record • Installation • Acknowledgement • Citation

Open Source Remote Sensing Self-Supervised Learning. The repository is OPEN, Pull requests are welcome!

TODO List

• Create Stable VISSL Version
• Simplify Installation Process
• Reduced Dependency Package
• Add Actions
• Complete CI/CD
• Simplify Pretrain Config File [Add PROC_ID]
• Complete the Pretrain Process
• Check the VISSL Pretrain
• Add Remote Sensing Image Segmentation Code
• Add Remote Sensing Image Classfication Code
• Add Remote Sensing Image Object Detection Code
• Organize the Code with LangChain Style
• Create Web Page
• Create stable OpenRSSSL Version
• Complete the Whole Process: From Pretrain to Specific-Task
• Make Stable Fairscale

Version Record

2024-03-03 openrsssl (alpha version)

Installation

1. Create Environment and Install Pytorch

conda create -n openrsssl_env python=3.8 -y
conda activate openrsssl_env
conda install pytorch==2.0.0 torchvision==0.15.0 torchaudio==2.0.0 pytorch-cuda=11.7 -c pytorch -c nvidia

2. Install OpenRSSSL

Download OpenRSSSL source code and switch to the source path for installation:

pip install --progress-bar off -r requirements.txt
pip install -e .[dev]

3. Optional: Install Apex from source

git clone --recursive https://www.github.com/NVIDIA/apex
cd apex
python3 setup.py install

Support Self-supervised Method

Supported Backbones	General Self-supervised Learning Method	Remote Sensing Self-supervised Learning Method
ResNet ResNeXt RegNet MobileNet Swin-Transformer EfficientNet ConvNeXt BEiT XCiT	Contrastive: SimCLR MoCo Barlow Twins SwAV Generative: MSN Self-Distillation: DINO (DeiT) iBOT (DeiT) Coming soon: BYOL * DenseCL * SimSiam * MAE * SimMIM * EVA * Deconstructing Denoising Diffusion Models for Self-Supervised Learning * DINO v2 Jigsaw	Coming soon: GLCNet (TGRS) * FALSE (GRSL) (coming soon.....) GraSS (TGRS) (coming soon.....)

Acknowledgement

We would like to thank the VISSL and MMSegmentation for its open-source project.