Information processing apparatus, information processing method, and storage medium

1 . An information processing apparatus comprising:

at least one processor; and

a memory coupled to the at least one processor, the memory storing instructions that, when executed by the at least one processor, cause the at least one processor to:

perform inference on an input using a first machine learning model; and

select a second machine learning model, in which at least some of components of a network structure of the first machine learning model have been changed, as a machine learning model used for the inference, in response to a predetermined condition being satisfied,

wherein the first machine learning model has a first component and a second component,

wherein the second machine learning model has at least a component in which a parameter or a path between nodes in the first component or the second component has been changed,

wherein the first machine learning model includes a first inference unit having a first parameter and a second parameter different from the first parameter, as the first component, and a second inference unit that takes an output of the first inference unit as an input, as the second component,

wherein the second machine learning model includes a third inference unit in which the second parameter in the first inference unit has been replaced with the first parameter, and a fourth inference unit having a same network structure as the second inference unit and taking an output of the third inference unit as an input,

wherein the first machine learning model includes a first inference unit that outputs a first output from first inference processing and a second output from second inference processing, as the first component, and a second inference unit that takes the first output and the second output as inputs, as the second component,

wherein the second machine learning model includes a fifth inference unit that outputs a first output from first inference processing, and a sixth inference unit that takes the first output and the second output that has been output previously as inputs,

wherein the inference is processing for restoring a degraded image that is input, and

wherein the degraded image is an image degraded by noise, compression, low resolution, blur, aberration, a defect, or a drop in contrast.

2 . The information processing apparatus according to claim 1 ,

wherein the predetermined condition is satisfied when an operation state of the information processing apparatus is a predetermined state.

3 . The information processing apparatus according to claim 2 ,

wherein the operation state is a usage state of a resource in the inference.

4 . The information processing apparatus according to claim 3 ,

wherein the instructions cause the at least one processor to:

switch the machine learning model used for the inference to the second machine learning model that is a machine learning model consuming fewer resources during the inference than the first machine learning model, when a usage rate of the resource is higher than a predetermined threshold for at least a predetermined percentage of a predetermined period.

5 . The information processing apparatus according to claim 3 ,

wherein the instructions cause the at least one processor to:

select the second machine learning model that has a lower number of layers than the first machine learning model, when a usage rate of the resource is higher than a predetermined threshold for at least a predetermined percentage of a predetermined period.

6 . The information processing apparatus according to claim 3 ,

wherein the instructions cause the at least one processor to:

obtain a user setting pertaining to the inference,

wherein the predetermined condition is satisfied when the user setting is a predetermined state.

7 . The information processing apparatus according to claim 6 ,

wherein the instructions cause the at least one processor to:

obtain, as the user setting, a setting as to whether to prioritize speed or accuracy in the inference,

wherein when speed is set to be prioritized in the inference, the machine learning model used for the inference is switched to the second machine learning model that is a machine learning model performing the inference faster than the first machine learning model, and when accuracy is set to be prioritized in the inference, the machine learning model is switched to the second machine learning model that is a machine learning model having a higher inference accuracy than the first machine learning model.

8 . The information processing apparatus according to claim 6 ,

wherein the inference is processing for restoring a degraded image that is input, and

the instructions cause the at least one processor to:

obtain, as the user setting, a setting for image quality in the restoring, and

take a case where the setting for the image quality in the restoring has been changed as the predetermined condition being satisfied, and switch the machine learning model used for the inference to the second machine learning model.

9 . The information processing apparatus according to claim 6 ,

wherein the inference is processing for restoring a degraded image that is input, and

the instructions cause the at least one processor to:

obtain, as the user setting, a setting for a degree of the restoring, and

take a case where the setting for the degree of the restoring has been changed as the predetermined condition being satisfied, and switch the machine learning model used for the inference to the second machine learning model.

10 . The information processing apparatus according to claim 6 ,

wherein the inference is processing for restoring a degraded image that is input, and

the instructions cause the at least one processor to:

obtain, as the user setting, a setting as to whether the input degraded image was captured indoors or captured outdoors, and

take a case where the setting as to whether the input degraded image was captured indoors or captured outdoors has been changed as the predetermined condition being satisfied, and switch the machine learning model used for the inference to the second machine learning model.

11 . The information processing apparatus according to claim 6 ,

wherein the inference is processing for restoring a degraded image that is input, and

the instructions cause the at least one processor to:

obtain, as the user setting, a setting pertaining to an image capturing apparatus that captured the input degraded image, and

take a case where the input degraded image was captured by a predetermined image capturing apparatus as the predetermined condition being satisfied, and switch the machine learning model used for the inference to the second machine learning model.

12 . The information processing apparatus according to claim 1 ,

wherein the inference is processing for classifying a subject in an image that is input.

13 . An information processing method comprising:

performing inference on an input using a first machine learning model; and

selecting a second machine learning model, in which at least some of components of a network structure of the first machine learning model have been changed, as a machine learning model used for the inference, in response to a predetermined condition being satisfied,

wherein the first machine learning model has a first component and a second component,

wherein the second machine learning model has at least a component in which a parameter or a path between nodes in the first component or the second component has been changed,

wherein the first machine learning model includes a first inference unit having a first parameter and a second parameter different from the first parameter, as the first component, and a second inference unit that takes an output of the first inference unit as an input, as the second component,

wherein the second machine learning model includes a third inference unit in which the second parameter in the first inference unit has been replaced with the first parameter, and a fourth inference unit having a same network structure as the second inference unit and taking an output of the third inference unit as an input,

wherein the first machine learning model includes a first inference unit that outputs a first output from first inference processing and a second output from second inference processing, as the first component, and a second inference unit that takes the first output and the second output as inputs, as the second component,

wherein the second machine learning model includes a fifth inference unit that outputs a first output from first inference processing, and a sixth inference unit that takes the first output and the second output that has been output previously as inputs,

wherein the inference is processing for restoring a degraded image that is input, and

wherein the degraded image is an image degraded by noise, compression, low resolution, blur, aberration, a defect, or a drop in contrast.

14 . A non-transitory computer readable storage medium storing program that, when executed by a computer causes the computer to perform an information processing method comprising:

performing inference on an input using a first machine learning model; and

selecting a second machine learning model, in which at least some of components of a network structure of the first machine learning model have been changed, as a machine learning model used for the inference, in response to a predetermined condition being satisfied,

wherein the first machine learning model has a first component and a second component,

wherein the second machine learning model has at least a component in which a parameter or a path between nodes in the first component or the second component has been changed,

wherein the first machine learning model includes a first inference unit having a first parameter and a second parameter different from the first parameter, as the first component, and a second inference unit that takes an output of the first inference unit as an input, as the second component,

wherein the second machine learning model includes a third inference unit in which the second parameter in the first inference unit has been replaced with the first parameter, and a fourth inference unit having a same network structure as the second inference unit and taking an output of the third inference unit as an input,

wherein the first machine learning model includes a first inference unit that outputs a first output from first inference processing and a second output from second inference processing, as the first component, and a second inference unit that takes the first output and the second output as inputs, as the second component,

wherein the second machine learning model includes a fifth inference unit that outputs a first output from first inference processing, and a sixth inference unit that takes the first output and the second output that has been output previously as inputs,

wherein the inference is processing for restoring a degraded image that is input, and

wherein the degraded image is an image degraded by noise, compression, low resolution, blur, aberration, a defect, or a drop in contrast.