Abstract: Few-shot learning aims to develop models with strong generalization capabilities using a small number of training samples. However, most learning methods rely solely on the visual features ...