GA-Net: Ghost convolution adaptive fusion skin lesion segmentation network

Longsong Zhou, Liming Liang, Xiaoqi Sheng

Abstract: Highlights•A GA-Net method is proposed to improve the diagnosis of skin diseases by automatically segmenting skin lesion regions.•A ghost module is designed to extract lesion area information, which generate rich feature maps to get more information.•The adaptive fusion module and bilateral attention module are designed to enhance the image semantic information delivery.