LB-UNet

PyTorch implementation for LB-UNet: A Lightweight Boundary-assisted UNet for Skin Lesion Segmentation (LB-UNet) (MICCAI 2024)

Network Architecture

Environments

1. torch 2.1.0

2. torchvision 0.16.0

3. python 3.8.0

Dataset

The dataset utilizes the preprocessed datasets provided by EGE-UNet from Google Drive.

The boundary maps that you can generate using boundary.py or download from Google Drive. ISIC2017, ISIC2018.

Code

1. Update the input and output paths for Boundary.py to align with the directory structure of dataset.py.

2. Run Boundary.py to generate boundary maps.

3. Update the data_path configuration in config_setting.py.

4. Run train.py to train the model.

Visualization