Siamese labels auxiliary learning

In deep learning, auxiliary modules for model training have become increasingly popular, such as Deep Mutual Learning (DML) and Multi-Scale Dense Convolutional Networks (MSDNet), which can maximize the performance of without increasing amount test computation. Nevertheless, current research does not fully exploit knowledge between different modules. This paper proposes a new technique–Siamese Labels Auxiliary (SiLA) Learning, in SiLA module is designed to concatenate outputs get information. addition, learned from be exploited by Our experiments show that effectively improves standard models achieves convincing experimental results on image classification tasks Moreover, easily combined with methods DML MSDNet perform best.