abstract |
The invention discloses a defect detection method based on a variational automatic encoder, which collects a product data set, divides the data set to obtain a training set and a test set, and generates image blocks scrambled according to labels S i from pictures in the training set; Build a variational auto-encoder network, and send the scrambled image blocks into the encoder network to obtain latent features; build a decoder network, input the latent features output from the encoder into the decoder, assist in solving the puzzle, and capture global and local information, thereby The high-resolution image is reconstructed, and the network parameters are updated by back-propagating the model with the preset loss function to obtain a trained model; the test is performed on the test set to complete the defect detection. The experimental results show the excellent generalization ability and defect detection ability of the model. |