Image evaluation apparatus and image evaluation method

An image evaluation apparatus that classifies defect information in a defect image of an electronic device by using a plurality of classifiers by machine learning, the image evaluation apparatus, comprising:

• a hardware processor;
• and
• a non-transitory computer-readable medium operatively coupled with the hardware processor, and storing the defect image of the electronic device and program instructions executed by the hardware processor that cause the image evaluation apparatus to:
• classify, by the plurality of classifiers, the defect information via the machine learning based on image-of-interest information of the defect image;
• extract a plurality of pieces of image-of-interest information;
• and
• evaluate the plurality of classifiers by comparing the plurality of pieces of image-of-interest information and determine an evaluation result including a center distance between a center of a first piece of the image-of-interest information that is created from a first viewpoint and a center of a second piece of the image-of-interest information, which is created from a second viewpoint that is different from the first viewpoint, or a rectangle multiplicity degree indicating a ratio of an overlap region, in which the first piece of the image-of-interest information and the second piece of the image-of-interest information overlap each other, to a total region.

An image evaluation method of classifying defect information in a defect image of an electronic device by using a classifier by machine learning, the image evaluation method, comprising:

• storing the defect image of the electronic device;
• storing defect region information in the defect image;
• classifying, by the classifier, the defect information based on image-of-interest information of the defect image;
• extracting the image-of-interest information;
• and
• evaluating classifiability of the defect image by comparing the image-of-interest information with the defect region information and determining an evaluation result including a center distance between a center of the image-of-interest and a center of the defect region, or a rectangle multiplicity degree indicating a ratio of an overlap region, in which the region of interest and the defect region overlap each other, to a total region.

An image evaluation method of classifying defect region information in a defect image of an electronic device by using a plurality of classifiers by machine learning, the image evaluation method, comprising:

• storing the defect image of the electronic device;
• classifying, by the plurality of classifiers, the defect region information based on image-of-interest information of the defect image;
• extracting a plurality of pieces of image information;
• and
• evaluating the plurality of classifiers by comparing the plurality of pieces of image-of-interest information and determining an evaluation result including a center distance between a center of a first piece of the image-of-interest information that is created from a first viewpoint and a center of a second piece of the image-of-interest information, which is created from a second viewpoint that is different from the first viewpoint, or a rectangle multiplicity degree indicating a ratio of an overlap region, in which the first piece of the image-of-interest information and the second piece of the image-of-interest information overlap each other, to a total region.