Figure 1. The overall structure of the method.

This study’s 5-shot image classification framework consists of the following steps: employing FResNet to extract image features and obtain feature vectors for support set samples and a query sample. The Adaptive Prototype module calculates the distances, denoted as d_i, between support set samples and the query sample, generating class prototypes accordingly. Finally, the Euclidean distance is utilized to measure the distance between the support set class prototypes and the query sample, facilitating the classification process. Figure source credit: CUB dataset.