Deep Residual Shrinkage Network For Few-Shot Learning

Abstract

The few-shot image classification is a challenging emerging direction in the field of deep learning, and it is widely used in oil well positioning, nondestructive testing in petroleum and other industrial fields. Aiming at the problem that traditional deep learning methods cannot effectively extract few-shot image features, this paper proposes a few-shot image classification algorithm based on deep residual shrinkage network (DRSN). On the one hand, to improve the ability of feature extraction, the attention mechanism is introduced into the residual network to construct the DRSN. On the other hand, we carry out calibration for small sample data to solve the problem of uneven data distribution and improve accuracy in the few-shot classification tasks. Experimental results show that the algorithm in the paper can perform effective image feature extraction, and have excellent performance in the few-shot classification tasks.