Match them up: visually explainable few-shot image classification

Abstract Few-shot learning (FSL) approaches, mostly neural network-based, assume that pre-trained knowledge can be obtained from base (seen) classes and transferred to novel (unseen) classes. However, the black-box nature of networks makes it difficult understand what is actually transferred, which may hamper FSL application in some risk-sensitive areas. In this paper, we reveal a new way perform for image classification, using visual representation backbone model patterns generated by self-attention based explainable module. The weighted only includes minimum number distinguishable features visualized serve as an informative hint on knowledge. On three mainstream datasets, experimental results prove proposed method enable satisfying explainability achieve high classification results. Code available at https://github.com/wbw520/MTUNet .