Mini-batch learning apparatus, operation program of mini-batch learning apparatus, operation method of mini-batch learning apparatus, and image processing apparatus

A non-transitory computer readable recording medium storing an operation program of a mini-batch learning apparatus that learns a machine learning model for performing semantic segmentation, which determines a plurality of classes in an image in units of pixels, by inputting mini-batch data to the machine learning model, the program causing a computer to function as:

• a calculation unit that calculates an area ratio of each of the plurality of classes in the mini-batch data;
• a specifying unit that specifies a rare class of which the area ratio is lower than a setting value;
• and
• an update unit that updates the machine learning model according to a result of the learning using the mini-batch data and sets an update level of the machine learning model in a case where the rare class is specified by the specifying unit to be lower than an update level of the machine learning model in a case where the rare class is not specified by the specifying unit.

An operation method of a mini-batch learning apparatus that learns a machine learning model for performing semantic segmentation, which determines a plurality of classes in an image in units of pixels, by inputting mini-batch data to the machine learning model, the method comprising steps of:

• calculating an area ratio of each of the plurality of classes in the mini-batch data;
• specifying a rare class of which the area ratio is lower than a setting value;
• and
• updating the machine learning model according to a result of the learning using the mini-batch data and setting an update level of the machine learning model in a case where the rare class is specified in the specifying step to be lower than an update level of the machine learning model in a case where the rare class is not specified in the specifying step.

An image processing apparatus that learns a machine learning model for performing semantic segmentation, which determines a plurality of classes in an image in units of pixels, by inputting mini-batch data to the machine learning model, the apparatus being configured to:

• calculate an area ratio of each of the plurality of classes in the mini-batch data;
• specify a rare class of which the area ratio is lower than a setting value;
• and
• set an update level of the machine learning model in a case where the rare class is specified to be lower than an update level of the machine learning model in a case where the rare class is not specified.