Training method, apparatus and non-transitory computer readable medium for a machine learning model

A training apparatus comprising a processor, the processor being configured to:

• compute a first feature map and a first attention map from a first image, and compute a second feature map and a second attention map from a second image, by inputting the first image, which does not include a defective area of a target, and the second image, which includes the defective area of the target, to a machine learning model;
• compute a first loss, based on the first attention map;
• compute a product of a multiplication between a value of each of elements of the second feature map and a value of each of elements of the second attention map as an integrated map by inputting the second feature map and the second attention map to the machine learning model;
• compute a class classification of the target by inputting the integrated map to the machine learning model;
• compute a second loss, based on the class classification;
• compute a total loss, based on the first loss and the second loss;
• and
• update a parameter of the machine learning model in such a manner as to minimize the total loss.

A non-transitory computer readable medium including computer executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform a method comprising:

• computing a first feature map and a first attention map from a first image, and computing a second feature map and a second attention map from a second image, by inputting the first image, which does not include a defective area of a target, and the second image, which includes the defective area of the target, to a machine learning model;
• computing a first loss, based on the first attention map;
• computing a product of a multiplication between a value of each of elements of the second feature map and a value of each of elements of the second attention map as an integrated map by inputting the second feature map and the second attention map to the machine learning model;
• computing a class classification of the target by inputting the integrated map to the machine learning model;
• computing a second loss, based on the class classification;
• computing a total loss, based on the first loss and the second loss;
• and
• updating a parameter of the machine learning model in such a manner as to minimize the total loss.