SAMCF: Adaptive global style alignment and multi-color spaces fusion for joint optic cup and disc segmentation

Longjun Huang, Ningyi Zhang, Yugen Yi, Wei Zhou, Bin Zhou, Jiangyan Dai, Jianzhong Wang

Abstract: Highlights•A simple and efficient DG method named AGSA is designed to generate a generic representation to avoid domain shift problem.•RGB, HSV and LAB are introduced to tackle the problem of limited visual representation and brightness sensitivity.•A BA submodule and a SSC submodule are proposed to integrate the advantages between high-level and low-level features.•An Edge Aware Loss is introduced to help model more focus on pixels close to edge of OD and OC, rather than other pixels.