Federal learning-based surface defect detection model training method

Abstract

The invention discloses a surface defect detection model training method based on federated learning, and relates to the field of federated learning, the method comprises the following steps: 1, client local data processing; step 2, the client pre-selects an algorithm; step 3, performing local training on the client; step 4, performing a parameter aggregation algorithm based on image quality weight; and 5, model combination. According to the method, the period of collecting the surface defect pictures is shorter, local data of the training participants cannot be leaked, the influence of a temporary poor communication condition on the overall training speed is reduced, and meanwhile the influence of the training participants with low-quality pictures on the overall model performance is also reduced.

Description

technical field [0001] The invention relates to the field of federated learning, in particular to a method for training a surface defect detection model based on federated learning. Background technique [0002] In the production process of industrial products, due to the production process or human factors, there may be scratches, cracks, holes and other defects on the surface of the products. These defects greatly reduce the quality of industrial products. Therefore, these products need to be checked in the quality inspection process. filter out. Surface defect detection refers to the use of computer programs to automatically determine whether there are defects on the surface of a product. Compared with manual quality inspection, surface defect detection technology is more efficient and is not affected by fatigue working conditions. The current surface defect detection is generally based on machine vision methods, that is, a neural network model (which can be called a sur...

AI Technical Summary

Problems solved by technology

[0007] For the first method, although various random processes have been performed on the picture, the structural features of the picture have changed, but the features of the defect have not changed, so the defect features for the model to learn are still insufficient.

For the second method, although the model can be trained, due to the small number of defective pictures, the final trained model will reach an over-fitting state, the generalization performance is low, and it cannot be put into use

For the third method, although it is technically feasible, in real life, the product surface pictures held by various manufacturers contain commercial secrets such as production technology and packaging technology, so manufacturers may not agree to conduct centralized training, which is less feasible. Low

Images