Semi-supervised self-distillation processing method and device based on noise and unmarked data

The invention relates to a semi-supervised self-distillation processing method and device based on noise and unmarked data. The method comprises the following steps: acquiring an original data set; extracting a marked and clean first data set from the original data set, and preprocessing the first data set to obtain a preprocessed first data set; training a data classification model by using the first data set to obtain a pre-trained classification model; marking and preprocessing unmarked data in the original data set by using a pre-training classification model to obtain a preprocessed second data set; training a data classification model by using the second data set to obtain a semi-supervised training classification model; and performing self-distillation training of the data classification model by using the first data set and the second data set to obtain a self-distillation training classification model. According to the technical scheme, the collected data are effectively utilized, and the precision o