Auxiliary Attribute Aided Few-shot Representation Learning for Gun Image Retrieval

Few-shot representation learning is one of the most challenging tasks in machine learning research field. The related applications including gun image retrieval usually achieve limited performance due to the lack of learning samples. In this paper, We propose a flexible and conceptually straightforward framework for few-shot gun image retrieval. We use ResNet as backbone network and design a hierarchical loss system based on auxiliary attributes extracted from different layers. Enhanced by a series of auxiliary attributes, discriminative features are learned efficiently. Experiments on a gun image dataset demonstrate the effectiveness of the proposed approach. In addition, it is worth noting that our framework can be easily extended to other few-shot learning tasks.