基于交叉自监督和DWI的NIID智能诊断方法

曹飞, 徐芊芊, 陈浩, 祖洁, 李晓文, 田锦, 鲍磊

An Intelligent Diagnosis Method for NIID Based on Cross Self-supervision and DWI

CAO Fei, XU Qianqian, CHEN Hao, ZU Jie, LI Xiaowen, TIAN Jin, BAO Lei

图3 交叉自监督学习架构. A为网络整体架构，B为ResNet50主体结构，C为ViT主体结构. logits1为ResNet50的输出，logits2为ViT的输出，Conv为卷积操作，Norm为归一化层

Fig. 3 Architecture of cross self-supervision. The part of A shows the overall network structure, while the part of B displays the main structure of ResNet50, the part of C displays the main structure of ViT. logits1 is the output of ResNet50, logits2 is the output of ViT, Conv is the convolution operation, and Norm is the normalization layer