Image data processing method for improving information multiplexing based on defect detection scene

The invention discloses an image data processing method for improving information multiplexing based on a defect detection scene, the technical combination of a cross entropy function, a convolutional neural network architecture and the like is adopted, and the convolutional neural network architecture uses a YOLOv5 architecture and mainly comprises a Focus module, a CBS module and the like. According to the method, a SiLU activation function, batch standardization and a Sigmoid function are adopted for optimization, so that a surface defect detection model with better performance is trained, meanwhile, rapid calculation can be carried out, the training time is shortened, the generated model can be used for carrying out real-time detection on a surface defect target in a complex industrial scene, and the detection efficiency is improved. According to the method, the detection network can pay more attention to strengthening the relation between the space information and the channel information, the adverse eff