【摘要】 为了提升关系网络图像分类的准确度，在网络中引入多尺度通道注意力机制，提出了一种新的小样本图像分类算法。由于多尺度通道注意力机制能够关注样本特征空间的重要信息，该方法能够提取图像更丰富的多尺度特征，并通过关系度量，改善了分类结果。实验结果表明，在MiniImageNet和Omniglot数据集上，该算法对图像分类精度有明显的提高。更多还原

【Abstract】 In order to improve the accuracy of image classification, a multi-scale channel attention is introduced into Relation Network, and a new few-shot image classification algorithm is proposed. Since the multi-scale channel attention can focus on important information in the sample feature space, this method can extract richer multi-scale features of the image, and improve the classification results through relationship measurement. Experimental results show that the proposed algorithm can improve the accuracy of image classification on Mini Image Net and Omniglot data sets.更多还原