CAWANet

<img src="https://github.com/ZGWzzu/CAWANet/blob/main/docs/CAWANet_.jpg" width = 800px>

Dataset

We used data sets adopted by FDSNet: MSD,NEU-Seg, MSD.

Environment

The training was on an NVIDIA 3090 GPU, and the FPS was tested on an NVIDIA 1080Ti GPU
Python 3.8.18 PyTorch 2.1.2 CUDA 11.8

einops==0.7.0
numpy==1.24.4
Pillow==10.0.1
torch==2.1.2
torchvision==0.16.2

You can create an environment by typing a command：

conda create -n CAWANet python==3.8.18
activate CAWANet

Download the required package:

pip install -r requirements.txt

Usage

We divided the test code for the three data sets into test_NEU.py, test_MSD.py, and test_MT.py. We set the parameters for each data set in the corresponding py file and just do the following to run it:

1. Download the appropriate data set and place the data set in the /CAWANet/ directory.
2. Download the weight files for the corresponding data set, and then place the weight files in the /CAWANet/ directory.
3. Run python test_x.py

Model weight

Dataset | Baidu Cloud/pth |Google Drive/pth | mIoU | FPS --- | --- | ---|---|--- MSD | CAWA_MSD Extract code：8bnv |CAWA_MSD| 90.1 | 186.5 NEU | CAWA_NEU Extract code：hu1s | CAWA_NEU|77.9 | 420.0 MT | CAWA_MT Extract code：3qtq|CAWA_MT| 79.5 | 414.7

Results

<img src="https://github.com/ZGWzzu/CAWANet/blob/main/docs/MSD222.jpg" width = 800px> <img src="https://github.com/ZGWzzu/CAWANet/blob/main/docs/neu111.jpg" width = 800px> <img src="https://github.com/ZGWzzu/CAWANet/blob/main/docs/MT_SORT.jpg" width = 800px>