基于人体红外数据的小样本蒸馏网络分类算法

In order to solve the problem of having too few samples in infrared thermal imaging data of metabolic syndrome, a distillation network based auxiliary diagnosis model for metabolic syndrome is proposed by combining knowledge distillation and few-shot learning methods. Compared with traditional diagnosis methods, the model can improve the diagnosis results while avoiding overfitting caused by small sample size. At the same time, the model is compressed by inheriting some parameters of the teacher network through the student network to reduce the calculation caused by the excessive data of each sample.