鉴于Transformer的Self-Attention机制具有优秀的表征能力,许多研究者提出了基于Self-Attention机制的图像处理模型,并取得了巨大成功.然而,基于Self-Attention的传统图像分类网络无法兼顾全局信息和计算复杂度,限制了 Self-Atten-tion的广泛应用.文中提出了 一种有效的、可扩展的注意力模块Local Neighbor Global Self-Attention(LNG-SA),该模块在任意时期都能进行局部信息、邻居信息和全局信息的交互.通过重复级联LNG-SA模块,设计了一个全新的网络,称为LNG-Transformer.该网络整体采用层次化结构,具有优秀的灵活性,其计算复杂度与图像分辨率呈线性关系.LNG-SA模块的特性使得LNG-T ransformer即使在早期的高分辨率阶段,也可以进行局部信息、邻居信息和全局信息的交互,从而带来更高的效率、更强的学习能力.实验结果表明,LNG-Transformer在图像分类任务中具有良好的性能.
LNG-Transformer:An Image Classification Network Based on Multi-scale Information Interaction
Due to the superior representation capability of the Transformer's Self-Attention mechanism,several researchers have developed Self-Attention mechanism-based image processing model and achieved great success.However,the traditional network for image classification based on Self-Attention cannot take into account global information and computational complexity,which limits the wide application of Self-Attention.This paper proposes an efficient and scalable attention module,Local Neighbor Glo-bal Self-Attention(LNG-SA),that may interact with local,neighbor,and global information at any stage.By cascading LNG-SA module,a brand-new network called LNG-Transformer is created.LNG-Transformer adopts a hierarchical structure that provides excellent flexibility,and has a computational complexity proportional to image resolution.The features of LNG-SA enable LNG-Transformer to interact with local information,neighbor information,and global information even in the early stage of high-reso-lution,resulting in increased efficiency and enhanced learning capacity.Experimental results show that LNG-Transformer per-forms well at image classification.
万方数据
维普
