Storage medium storing program, image processing apparatus, and training method of machine learning model

An image processing apparatus comprising:

• a controller;
• and
• a memory storing instructions, the instructions, when executed by the controller, causing the image processing apparatus to:
• acquire input image data indicating an input image;
• input the input image data to a trained machine learning model configured to perform calculation processing on the input image data and to generate output image data corresponding to the input image data, the machine learning model being trained to generate the output image data, the machine learning model being trained by:
• acquiring training input image data indicating a training input image formed by a plurality of pixels;
• acquiring teacher image data corresponding to the training input image data, the teacher image data including first teacher image data and second teacher image data, the first teacher image data representing a first teacher image, the first teacher image being acquired by performing first image processing in a first area of the training input image, the first teacher image having vertical and horizontal sizes same as the training input image, the first teacher image data including a first parameter indicating a level of the first image processing for each of the plurality of pixels, the first parameter of each pixel in the first area being larger than zero, the first parameter of each pixel other than the first area being zero, the second teacher image data representing a second teacher image, the second teacher image being acquired by performing second image processing in a second area of the training input image, the second teacher image having vertical and horizontal sizes same as the training input image, the second teacher image data including a second parameter indicating a level of the second image processing for each of the plurality of pixels, the second parameter of each pixel in the second area being larger than zero, the second parameter of each pixel other than the second area being zero, the first area being an area in which it is preferable to perform the first image processing, the second area being an area in which it is preferable to perform the second image processing different from the first image processing;
• and
• adjusting a plurality of parameters of the machine learning model by using a plurality of sets of data pairs each including the training input image data and the teacher image data corresponding to the training input image data;
• perform the first image processing on first data out of the input image data, the first data indicating the first area identified by the output image data;
• and
• perform the second image processing on second data out of the input image data, the second data indicating the second area identified by the output image data.

A training method of a machine learning model configured to perform calculation processing on input image data indicating an input image and to generate output image data corresponding to the input image data, the method comprising:

• acquiring training input image data indicating a training input image formed by a plurality of pixels;
• acquiring teacher image data corresponding to the training input image data, the teacher image data including first teacher image data and second teacher image data, the first teacher image data representing a first teacher image, the first teacher image being acquired by performing first image processing in a first area of the training input image, the first teacher image having vertical and horizontal sizes same as the training input image, the first teacher image data including a first parameter indicating a level of the first image processing for each of the plurality of pixels, the first parameter of each pixel in the first area being larger than zero, the first parameter of each pixel other than the first area being zero, the second teacher image data representing a second teacher image, the second teacher image being acquired by performing second image processing in a second area of the training input image, the second teacher image having vertical and horizontal sizes same as the training input image, the second teacher image data including a second parameter indicating a level of the second image processing for each of the plurality of pixels, the second parameter of each pixel in the second area being larger than zero, the second parameter of each pixel other than the second area being zero, the first area being an area in which it is preferable to perform the first image processing, the second area being an area in which it is preferable to perform the second image processing different from the first image processing;
• and
• adjusting a plurality of parameters of the machine learning model by using a plurality of sets of data pairs each including the training input image data and the teacher image data corresponding to the training input image data.