Learning method, computer program, classifier, generator, and processing system

A learning method comprising

• using a machine learning model for image generation that is defined to, when any data is input, output specific image data and
• a machine learning model for classification that is defined to, when image data output from the machine learning model for image generation or other image data is input, output a classification of the input image data,
• using data for learning of classifier and/or data for learning of generator including both or either one of image data in which a component that is difficult to visually judge is left out and image data in which a noise component that is difficult to visually judge is added at a predetermined ratio, and
• learning the machine learning model for classification and/or the machine learning model for image generation by GAN (Generative Adversarial Networks).

A computer program that causes a computer to perform processes of:

• storing therein definition data of a machine learning model for classification that is defined to, when image data is input, output a classification of an image based on the image data;
• giving data for learning of classifier that includes both or either one of image data in which a component that is difficult to visually judge is left out and image data in which a noise component that is difficult to visually judge is added at a predetermined ratio, to a machine learning model for classification based on the definition data;
• and
• learning the machine learning model for classification as a classifier.

A computer program that causes a computer to perform processes of:

• storing therein definition data of a machine learning model for image generation that is defined to output specific image data when any data is input, and
• definition data of a machine learning model for classification that is defined to, when the image data output from the machine learning model for image generation or other image data is input, output a classification of an image based on the input image data;
• performing both or either one of a process of leaving out a component that is difficult to judge visually and a process of adding a noise component that is difficult to judge visually for a predetermined ratio of the image data output from the machine learning model for image generation or the image data given to the machine learning model for classification;
• and
• learning the machine learning model for classification and the machine learning model for image generation by GAN.

A classifier comprising:

• an input unit to which image data is input;
• an output unit that outputs a classification of an image based on the image data;
• and
• a processing execution unit that executes a predetermined process for the image data by using a machine learning model, wherein
• the classifier is learned with training data including data for learning of classifier that includes both or either one of image data in which a component that is difficult to visually judge is left out and image data in which a noise component that is difficult to visually judge is added at a predetermined ratio, and a classification label of the image data included in the data for learning of classifier, and outputs a classification of input image data to a computer.

A generator comprising:

• an input unit to which any data is input;
• an output unit that outputs image data of an image generated based on the input data;
• and
• a processing execution unit that executes a predetermined process for the input data by using a machine learning model, wherein
• the generator is learned by using data for learning of generator that includes both or either one of image data obtained by leaving out a component that is difficult to visually judge in a predetermined ratio of the image data output from the output unit or image data given to the machine learning model, and image data obtained by adding a noise component that is difficult to visually judge in the predetermined ratio of the image data, and
• the generator outputs image data based on any data input thereto to a computer.