Display panel microdefect classification method based on multi-scale twin neural network

The invention relates to the technical field of display panel defect detection, and discloses a display panel microdefect classification method based on a multi-scale twin neural network, and the classification method comprises the following steps: (1) obtaining a data set; (2) constructing a direct classification model; (3) training a direct classification model on the large sample data set; (4) the model structure is optimized until it is judged that correct features of various types are learned through a visual convolution feature method; (5) constructing a multi-scale twin neural network by taking the feature extraction network of the direct classification model as a branch; (6) dividing the panel defect data set into a training set and a test set; (7) training the multi-scale twin neural network; and (8) comparing the similarity between the feature vector of the to-be-tested sample obtained through the feature extraction network and the mean vector of each type of sample, wherein the type with the highes