Fusion of global and adaptive local information for few-shot image classification

Ting Xiao, Yiqing Xia, Ruiqi Tang, Wenli Du, Zhe Wang

Abstract: Highlights•Novel FGAL method enhances bidirectional knowledge in global and local views.•Introduce classifier re-weighting optimizer and global–local knowledge collaboration.•Improve capability to capture overall features and image details.•Demonstrate effectiveness on few-shot and adaptability in cross-domain datasets.