Model training method and device, image classification method and device and readable storage medium

The invention provides a model training method, an image classification method, an image classification device and a computer readable storage medium. The model training method comprises the steps of obtaining a to-be-trained image, and extracting a feature matrix of the to-be-trained image; based on the feature matrix, obtaining a parameter matrix obtained by training; determining an abnormal value in the parameter matrix value; and training a classifier of the to-be-trained model by using the abnormal value. Through the above mode, the image classification device obtains the weight parameter matrix of model training by using an abnormal value detection method, unimportant and redundant parameters are rejected from the model, and parameters containing more effective information, namely parameters corresponding to abnormal values, are reserved, so that the model obtained through training is simpler, the training time is short, and the effect is remarkable. 本申请提供一种模型训练方法、图像分类方法、图像分类装置以及计算机可读存储介质。该模型训练方法包括：获取待训