Skip to content

readme_en.md

Latest commit

 

History

History
174 lines (137 loc) · 6.79 KB

readme_en.md

File metadata and controls

174 lines (137 loc) · 6.79 KB

Hyperspectral Image Classification (HSIC)

中文 (readme.md)

Project Overview & Features

This project focuses on hyperspectral image classification (HSI Classification), implementing a variety of mainstream and cutting-edge deep learning models (CNN, Transformer, Mamba, etc.) for remote sensing HSI analysis and classification. It supports multiple datasets and model comparison experiments, making it suitable for both research and engineering applications.

Key Features:

  • Supports various mainstream and novel deep learning models
  • Clean code structure, easy to extend
  • Multi-dataset and batch experiment support
  • Visualization and comparative analysis of results

Note: This project is a graduation design. Some implementations are for reference only. Feedback and suggestions are welcome.

1. Usage

  1. Clone the repository and install dependencies:

    git clone https://github.com/751K/HS_Image_Classification.git
cd HS_Image_Classification
pip install -r requirements.txt

  2. Run the main program:

    python src/main.py

  3. Train a specific model:

    cd src
python main.py

  4. Batch train all models:

    cd src
python run_all.py

  5. Visualize results:

    cd src
python vis.py

2. Supported Models & Paths

Model Name Type Path
ResNet2D CNN src.CNNBase.ResNet2D
ResNet3D CNN src.CNNBase.ResNet3D
HybridSN CNN src.CNNBase.HybridSN
LeeEtAl3D CNN src.CNNBase.LeeEtAl3D
GCN2D GCN src.CNNBase.GCN2D
SSFTT Transformer src.TransformerBase.SSFTT
SwimTransformer Transformer src.TransformerBase.SwimTransformer
VisionTransformer Transformer src.TransformerBase.VisionTransformer
SFT Transformer src.TransformerBase.SFT
SSMamba Mamba src.MambaBase.SSMamba
MambaHSI Mamba src.MambaBase.MambaHSI
STMamba Mamba src.MambaBase.STMamba
AllinMamba Mamba src.MambaBase.AllinMamba

Implementations are for reference only. Please adjust details and parameters according to papers and actual needs.

Model Extension

To add a new model:

  1. Add your model implementation under src/CNNBase/, src/TransformerBase/, or src/MambaBase/.
  2. Import the new model in the corresponding __init__.py.
  3. Configure the new model in src/config.py.

3. Project Structure

project/
├── datasets/           # Datasets
│   ├── Indian/         # Indian Pines
│   ├── Pavia/          # Pavia University
│   └── Salinas/        # Salinas
├── results/            # Model results and outputs
├── src/                # Source code
│   ├── CNNBase/        # CNN models
│   ├── MambaBase/      # Mamba models
│   ├── TransformerBase/# Transformer models
│   ├── Dim/            # Dimensionality reduction
│   ├── Train_and_Eval/ # Training and evaluation
│   ├── datesets/       # Data processing
│   ├── draw/           # Visualization
│   ├── config.py       # Config file
│   ├── main.py         # Main entry
│   ├── model_init.py   # Model initialization
│   └── run_all.py      # Batch training script
├── LICENSE             # License
└── requirements.txt    # Dependencies

4. Supported Datasets

  • Indian Pines
  • Pavia University
  • Salinas
  • KSC
  • Botswana
  • WHU-Hi-Honghu

Dataset Extension

To add a new dataset:

  1. Add loading logic in src/datesets/datasets_load.py.
  2. Configure the new dataset in src/config.py.

5. Experimental Results

Model Name Dataset Accuracy AA Kappa Train Time(s) Params Inference Time(s)
LeeEtAl3D Indian Pines 0.8022 0.7257 0.7703 119.60 158,736 0.4582
ResNet3D Indian Pines 0.9470 0.8830 0.9395 130.47 418,768 0.3068
ResNet2D Indian Pines 0.9871 0.9898 0.9853 559.36 11,193,744 1.6913
GCN2D Indian Pines 0.9026 0.6905 0.8887 22.85 23,441 0.0570
HybridSN Indian Pines 0.9500 0.9231 0.9429 92.43 10,885,248 0.2625
SwimTransformer Indian Pines 0.9461 0.8434 0.9385 81.72 53,264 0.2163
VisionTransformer Indian Pines 0.9646 0.9223 0.9596 83.04 117,264 0.2238
SSFTT Indian Pines 0.9376 0.8122 0.9289 20.24 165,536 0.0287
SFT Indian Pines 0.8323 0.6128 0.8077 13.16 136,112 0.0197
MambaHSI Indian Pines 0.9429 0.8431 0.9348 862.80 105,232 2.7553
SSMamba Indian Pines 0.9648 0.9598 0.9599 289.79 240,050 0.7452
STMamba Indian Pines 0.8473 0.6145 0.8237 602.23 69,456 1.0509
AllinMamba Indian Pines 0.9879 0.9752 0.9862 417.85 2,232,084 1.3174

6. Environment

  • Supports Linux, MacOS, Windows (WSL2 recommended)
  • Python 3.12+

Dependencies

  1. PyTorch 2.1.0+
  2. Numpy
  3. scikit-learn
  4. See requirements.txt for details

Hardware Recommendation

  1. NVIDIA GPU (RTX 3060 or above recommended)
  2. At least 16GB RAM
  3. Apple Silicon is also supported, but NVIDIA GPU is recommended

7. FAQ

Q: Error or missing dependencies?
A: Please make sure all dependencies in requirements.txt are installed.

Q: How to customize models or datasets?
A: See the "Model Extension" and "Dataset Extension" sections.

Q: Can I run on CPU?
A: Yes, but not recommended.

8. Contribution

Issues, PRs, and suggestions are welcome! Please submit issues on GitHub for bugs or improvements.

9. Contact

10. License

This project is licensed under the MIT License. See LICENSE for details.

11. Acknowledgement

Thanks to myself and all future contributors. Thanks to ChatGPT and Claude for their assistance.