Glass container defect detection network lightweight method and system based on knowledge distillation

The invention belongs to the field of visual defect detection, and provides a glass container defect detection network lightweight method and system based on knowledge distillation, and the method comprises the steps: carrying out defect category screening, labeling and data preprocessing on a glass container defect picture; the method comprises the following steps: establishing a confidence distillation branch, training a teacher model to obtain a teacher model weight, reasoning the trained teacher model, extracting a plurality of prediction frames which have the highest confidence and are not overlapped from the teacher model, and training a student model to perform confidence distillation; establishing a global distillation branch, training a teacher model to obtain a teacher model weight, reasoning the trained teacher model, extracting multi-scale features from the teacher model, reloading the multi-scale features and the trained student model weight, and carrying out global distillation; the confidence d