File size: 3,441 Bytes
e2a3312 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 |
## 🛠️ Requirements
### Environment
- **Python** 3.8+
- **PyTorch** 1.13.0+
- **CUDA** 11.6+
- **Ubuntu** 18.04 or higher / Windows 10
### Installation
```bash
# Create conda environment
conda create -n dccs python=3.8 -y
conda activate dccs
# Install PyTorch
pip install torch==1.13.0 torchvision==0.14.0 torchaudio==0.13.0
# Install dependencies
pip install packaging
pip install timm==0.4.12
pip install pytest chardet yacs termcolor
pip install submitit tensorboardX
pip install triton==2.0.0
pip install causal_conv1d==1.0.0
pip install mamba_ssm==1.0.1
# Or simply run
pip install -r requirements.txt
```
## 📁 Dataset Preparation
We evaluate our method on three public datasets: **IRSTD-1K**, **NUAA-SIRST**, and **SIRST-Aug**.
| Dataset | Link |
|---------|------|
| IRSTD-1K | [Download](https://github.com/RuiZhang97/ISNet) |
| NUAA-SIRST | [Download](https://github.com/YimianDai/sirst) |
| SIRST-Aug | [Download](https://github.com/Tianfang-Zhang/AGPCNet) |
Please organize the datasets as follows:
```
├── dataset/
│ ├── IRSTD-1K/
│ │ ├── images/
│ │ │ ├── XDU514png
│ │ │ ├── XDU646.png
│ │ │ └── ...
│ │ ├── masks/
│ │ │ ├── XDU514.png
│ │ │ ├── XDU646.png
│ │ │ └── ...
│ │ └── trainval.txt
│ │ └── test.txt
│ ├── NUAA-SIRST/
│ │ └── ...
│ └── SIRST-Aug/
│ └── ...
```
## 🚀 Training
```bash
python main.py --dataset-dir '/path/to/dataset' \
--batch-size 4 \
--epochs 400 \
--lr 0.05 \
--mode 'train'
```
**Example:**
```bash
python main.py --dataset-dir './dataset/IRSTD-1K' --batch-size 4 --epochs 400 --lr 0.05 --mode 'train'
```
## 📊 Testing
```bash
python main.py --dataset-dir '/path/to/dataset' \
--batch-size 4 \
--mode 'test' \
--weight-path '/path/to/weight.tar'
```
**Example:**
```bash
python main.py --dataset-dir './dataset/IRSTD-1K' --batch-size 4 --mode 'test' --weight-path './weight/irstd1k_weight.pkl'
```
## 📈 Results
### Quantitative Results
| Dataset | IoU (×10⁻²) | Pd (×10⁻²) | Fa (×10⁻⁶) | Weights |
|:-------:|:------------:|:----------:|:----------:|:-------:|
| IRSTD-1K | 69.64 | 95.58 | 10.48 | [Download](https://drive.google.com/file/d/1KqlOVWIktfrBrntzr53z1eGnrzjWCWSe/view?usp=sharing) |
| NUAA-SIRST | 78.65 | 78.65 | 2.48 | [Download](https://drive.google.com/file/d/13JQ3V5xhXUcvy6h3opKs15gseuaoKrSQ/view?usp=sharing) |
| SIRST-Aug | 75.57 | 98.90 | 33.46 | [Download](https://drive.google.com/file/d/1lcmTgft0LStM7ABWDIMRHTkcOv95p9LO/view?usp=sharing) |
## 📂 Project Structure
```
DCCS/
├── dataset/ # Dataset loading and preprocessing
├── model/ # Network architecture
├── utils/ # Utility functions
├── weight/ # Pretrained weights
├── main.py # Main entry point
├── requirements.txt # Dependencies
└── README.md
```
## 🙏 Acknowledgement
We sincerely thank the following works for their contributions:
- [BasicIRSTD](https://github.com/XinyiYing/BasicIRSTD) - A comprehensive toolbox
- [MSHNet](https://github.com/ying-fu/MSHNet) - Scale and Location Sensitive Loss |