A learnable support selection scheme for boosting few-shot segmentation

Wenxuan Shao, Hao Qi, Xinghui Dong

Abstract: Highlights•Proposing to use a similarity-based support selection scheme to boost few-shot segmentation.•Adopting a Siamese Support Selection Network which can be inserted to a few-shot segmentation network.•Exploiting the local and global features extracted using a CNN and a Transformer network.