EchoSixHIYA/CAFNet

Python

CAFNet: Circular Attention for Medical Image Segmentation

CAFNet applies circular attention mechanisms to enhance the segmentation of medical images.

Experimental Setup (Using CVC-ClinicDB as an Example)

0. Environment Setup

Ensure all dependencies in requirements.txt are met. It is recommended to create a virtual environment using the requirements.txt.

1. Data Preparation

Download the original dataset and extract it to the ./data folder.
Run random.division.py to randomly split the dataset into training, validation, and test sets.
Execute process.py to preprocess all data. This will generate data_{train, val, test}.npy and mask_{train, val, test}.npy.

2. Training

Download the DeiT-base model from the DeiT repository and save it to the ./pretrained folder.
Download the ResNet-34 model from the timm PyTorch repository and save it to the ./pretrained folder.
Run train.py. You may need to adjust some parameters based on your specific settings.

3. Testing

Run test.py with the following command: python test.py --ckpt_path='path_to_your_check_point.pth