Defect detection model training method and device, equipment and storage medium

Abstract

The invention provides a flaw detection model training method and device, equipment and a storage medium, and the method comprises the steps: obtaining a first training sample image with flaw labeling information and a second training sample image without flaw labeling information, respectively inputting the first training sample image and the second training sample image into a flaw detection model serving as a student model to obtain a first flaw detection result and a second flaw detection result, and inputting the second training sample image into the teacher model to obtain a third flaw detection result indicating the flaw position and the prediction score of the target flaw category. And if the prediction score is greater than a credible threshold value of the target flaw category, determining that a third flaw detection result is credible, and taking the third flaw detection result as flaw labeling information of the second training sample image. And training a defect detection model according to respective defect detection results and defect labeling information of the first training sample image and the second training sample image. And a defect detection model with good performance is obtained in a semi-supervised training mode.

Description

technical field [0001] The present invention relates to the technical field of image processing, and in particular, to a method, device, equipment and storage medium for training a flaw detection model. Background technique [0002] Industrial defect (or defect) detection is a challenging task, and its application scenarios cover various application fields such as heavy industry and light industry. For example, industries such as steel, electronics, and textiles all have the need to detect surface defects of products. [0003] At present, the commonly used method is to use the defect detection model to complete the defect detection task of the product. The traditional defect detection model is trained based on the collected training samples with defect annotation information. However, the training sample data with flawed annotation information that can be collected on the industrial production line is also limited, which makes the performance of the trained model poor. SUM...

AI Technical Summary

Problems solved by technology

However, the training sample data with defect annotation information that can be collected on the industrial production line is also relatively limited, making the trained model perform poorly

Images