Image scoring apparatus, image scoring method and method of machine learning

1 . An information-processing device, comprising:

a Central Processing Unit (“CPU”); and

a memory storing instructions for causing the information-processing device, when executed by the CPU, to:

include at least a machine learning model configured to receive an input image and an attribute as input, and to output at least one region in the input image and an evaluation value for each of the at least one region,

wherein, for a common input image, the at least one region and the evaluation value output when one attribute is given are different from the at least one region and the evaluation value output when another attribute different from the one attribute is given,

wherein the instructions further cause the information-processing device to:

include, as the machine learning model, at least a heatmap output model configured to receive the input image and the attribute as input, and to output a heatmap corresponding to the attribute,

receive the heatmap as input, and to extract at least one region in the heatmap, and

receive a partial image of the input image corresponding to the at least one region as input, and to estimate the evaluation value for the partial image,

wherein the heatmap output model is a heatmap output model obtained by causing the heatmap output model to learn through use of teacher data which includes, as a question, the input image and the attribute, and includes, as an answer, the heatmap obtained by combining, in accordance with the attribute, a plurality of intermediate heatmaps obtained from a plurality of learned machine learning models which are different from one another, and generate intermediate heatmaps for the input of the input image.

2 . The information-processing device according to claim 1 ,

wherein the instructions further cause the information-processing device to:

further include, as the machine learning model, at least a region evaluation value output model configured to receive the input image and the heatmap as input, and to output the at least one region in the input image and the evaluation value for each of the at least one region,

wherein the region evaluation value output model is a region evaluation value output model obtained by causing the region evaluation value output model to learn through use of teacher data which includes, as a question, the heatmap and includes, as an answer, a region obtained from a learned machine learning model or an algorithm which outputs the region for input of the heatmap and the evaluation value obtained from a learned machine learning model which outputs the evaluation value for a portion of the input image corresponding to the region.

3 . An information-processing device, comprising:

a Central Processing Unit (“CPU”); and

a memory storing instructions for causing the information-processing device, when executed by the CPU, to:

include at least a machine learning model configured to receive an input image and an attribute as input, and to output at least one region in the input image and an evaluation value for each of the at least one region,

wherein, for a common input image, the at least one region and the evaluation value output when one attribute is given are different from the at least one region and the evaluation value output when another attribute different from the one attribute is given,

wherein the instructions further cause the information-processing device to:

include a single machine learning model as the machine learning model, wherein the single machine learning model is a machine learning model obtained by causing the machine learning model to learn through use of teacher data which includes, as a question, the input image and the attribute and includes, as an answer, a region obtained by inputting, into a learned machine learning model or an algorithm which outputs the region for input of a heatmap, a heatmap obtained by combining, in accordance with the attribute, a plurality of intermediate heatmaps obtained from a plurality of learned machine learning models which are different from one another, and generate intermediate heatmaps for the input of the input image, and the evaluation value obtained from a learned machine learning model which outputs the evaluation value for a portion of the input image corresponding to the region.

4 . An information-processing method of causing a computer to execute:

training a machine learning model as a heatmap output model by causing the machine learning model to learn through use of teacher data which includes, as a question, an input image and an attribute, and includes, as an answer, a heatmap obtained by combining, in accordance with the attribute, a plurality of intermediate heatmaps obtained from a plurality of learned machine learning models which are different from one another, and generate intermediate heatmaps for input of the input image.

5 . The information-processing method according to claim 4 , the method further causing a computer to execute:

training a second machine learning model as a region evaluation value output model by causing the second machine learning model to learn through use of teacher data which includes, as the question, the heatmap and includes, as the answer, a region obtained from a learnedthe learned machine learning model or an algorithm which outputs the region for input of the heatmap and an evaluation value obtained from the learned machine learning model which outputs the evaluation value for a portion of the input image corresponding to the region.

6 . The information-processing device according claim 1 , wherein the plurality of learned machine learning models have been independently trained.