Image processing apparatus, and operation method and operating program thereof, operating apparatus, and operation method and operating program thereof, and machine learning system
An image processing apparatus includes an extraction unit that extracts, from among a plurality of designated regions in which labels of classes are designated, complicated regions which are regions of at least a part of the designated regions and are regions having relatively complicated contours, in an annotation image given as learning data to a machine learning model for performing semantic segmentation in which a plurality of the classes in an image are discriminated on a per-pixel basis, and a setting unit that sets additional labels for the complicated regions separately from original labels originally designated for the annotation image.
1. An image processing apparatus comprising:
a processor that is configured to:
extract, from among a plurality of designated regions in which labels of classes are designated, complicated regions which are regions of at least a part of the plurality of designated regions and are regions having zigzag contours, in an annotation image given as learning data to a machine learning model for performing semantic segmentation in which a plurality of the classes in an image are discriminated on a per-pixel basis; and
set additional labels for the complicated regions separately from original labels originally designated for the annotation image.
2. The image processing apparatus according to claim 1 , wherein the processor sets the same additional label for a plurality of the complicated regions extracted from the plurality of designated regions in which the original labels indicate the same class.
3. The image processing apparatus according to claim 1 , wherein the processor performs processing of extracting the complicated region for each of the plurality of designated regions.
4. The image processing apparatus according to claim 1 , wherein in a case where the processor extracts, as the complicated region, a boundary between a first designated region and a second designated region, the processor selectively sets one additional label related to any one of the first designated region or the second designated region for the complicated region of the boundary.
5. The image processing apparatus according to claim 1 , wherein the processor sets, for a small complicated region equal to or less than a set size among the complicated regions, the original label of one of the plurality of designated regions including the small complicated region.
6. An operating apparatus comprising a processor configured to:
input an input image to the machine learning model which is trained by using the annotation image for which the additional label is set in the image processing apparatus according to claim 1 as the learning data to cause the machine learning model to perform the semantic segmentation, and outputs an output image; and
replace the additional label of the output image with the original label based on label information storing a relationship between the original label and the additional label.
7. An operation method of an image processing apparatus, comprising:
extracting, from among a plurality of designated regions in which labels of classes are designated, each of the plurality of designated region being surrounded by a contour, complicated regions which are regions of at least a part of the plurality of designated regions and are regions having zigzag contours, in an annotation image given as learning data to a machine learning model for performing semantic segmentation in which a plurality of the classes in an image are discriminated on a per-pixel basis; and
setting additional labels for the complicated regions separately from original labels originally designated for the annotation image.
8. An operation method of an operating apparatus, comprising:
inputting an input image to a machine learning model which is trained by using the annotation image for which the additional label is set in the operation method of the image processing apparatus according to claim 7 as the learning data to cause the machine learning model to perform the semantic segmentation, and outputting an output image; and
replacing the additional label of the output image with the original label based on label information storing a relationship between the original label and the additional label.
9. A non-transitory computer-readable storage medium storing an operating program of an image processing apparatus causing a computer to:
extract, from among a plurality of designated regions in which labels of classes are designated, complicated regions which are regions of at least a part of the plurality of designated regions and are regions having zigzag contours, in an annotation image given as learning data to a machine learning model for performing semantic segmentation in which a plurality of the classes in an image are discriminated on a per-pixel basis; and
set additional labels for the complicated regions separately from original labels originally designated for the annotation image.
10. A non-transitory computer-readable storage medium storing an operating program of an operating apparatus causing a computer to:
input an input image to the machine learning model which is trained by using the annotation image for which the additional label is set in the operating program of the image processing apparatus according to claim 9 as the learning data to cause the machine learning model to perform the semantic segmentation, and outputs an output image; and
replace the additional label of the output image with the original label based on label information storing a relationship between the original label and the additional label.
11. A machine learning system comprising a processor configured to:
extract, from among a plurality of designated regions in which labels of classes are designated, complicated regions which are regions of at least a part of the plurality of designated regions and are regions having zigzag contours, in an annotation image given as learning data to a machine learning model for performing semantic segmentation in which a plurality of the classes in an image are discriminated on a per-pixel basis;
set additional labels for the complicated regions separately from original labels originally designated for the annotation image;
input an input image to the machine learning model which is trained by using the annotation image for which the additional label is set as the learning data to cause the machine learning model to perform the semantic segmentation, and outputs an output image; and
replace the additional label of the output image with the original label based on label information storing a relationship between the original label and the additional label.