A²Seg: Adversarial learning with channel-wise Attention for skin lesion Segmentation

Dependencies

• python 3.11 (below should be fine)
• Pytorch 2.0 (below should be fine)
• CUDA

Dataset

• ISBI International Skin Imaging Collaboration (ISIC) 2017 challenge, Part I Lesion Segmentation

Project Structure

File	Description
dataLoader.py	(as it says)
transform.py	Functions for data augmentation
segmentor.py	Model that predict segmentation masks, based on Attention U-net
critic.py	Model that takes masked images to produce multi-scale feature map
train.py	(as it says)

Model Structure

Results

References

• The code of this project is greatly inspired by SegAN (Yuan Xue, et al., 2018), sincere respects to their work and briliant ideas.