Multi-class identifier, method, and computer-readable recording medium

摘要

A multi-class identifier identifies a kind of an imager, and identifies in detail with respect to a specified kind of a group. The multi-class identifier includes: an identification fault counter providing the image for test that includes any of class labels to the kind identifiers so that the kind identifiers individually identify the kind of the provided image, and counting, for a combination of arbitrary number of kinds among the plurality of kinds, the number of times of incorrect determination in the arbitrary number of kinds that belongs to the combination; a grouping processor, for a group of the combination for which count result is equal to or greater than a predetermined threshold, adding a group label corresponding to the group to the image for learning that includes the class label corresponding to any of the arbitrary number of kinds that belongs to the group.

主权项

1. A multi-class identifier comprising:
a CPU which is configured to: perform a first learning processing of (i) receiving a plurality of first pieces of characteristic data for learning, each of the first pieces of characteristic data including a class label corresponding to one of a plurality of kinds, and (ii) generating kind identifiers based on the received first pieces of characteristic data, each of the kind identifiers individually identifying a kind of a respective one of the first pieces of characteristic data; perform an identification fault count processing of (i) providing a plurality of second pieces of characteristic data for testing to the kind identifiers so that each of the kind identifiers individually identifies a kind of a respective one of the provided second pieces of characteristic data, each of the second pieces of characteristic data including one of a plurality of class labels, and (ii) counting, for a combination of an arbitrary number of kinds from among the plurality of kinds, a number of times that an incorrect determination is made with respect to the kinds belonging to the combination; perform a grouping processing of adding, for a group of the combination for which a result of the counting is equal to or greater than a predetermined threshold, a group label corresponding to the group, the group label being added to pieces of the characteristic data for learning that include a class label corresponding to one of the arbitrary number of kinds belonging to the group; perform a second learning processing of (i) receiving the pieces of characteristic data for learning to which the group label has been added, (ii) generating a group identifier based on the received pieces of characteristic data that identifies the group of the received pieces of characteristic data, (iii) forming a first class identifier by combining the group identifier and a kind identifier that identifies a kind being excluded from the group, and (iv) generating an intra-group identifier that identifies an arbitrary number of kinds of characteristic data belonging to the group, to form a second class identifier; and generate an inter-class accumulating processor that (i) performs an identification judging processing according to an identification score value obtained by adding an identification score value of the first class identifier corresponding to the group, and a value in which an identification score value of the second class identifier is multiplied by a predetermined magnification, with respect to the group, and (ii) performs an identification judging processing according to the identification score value of the first class identifier with respect to the kind being excluded from the group.