Similarity calculating device and method, and recording medium

A method comprising, by a computer device,

• by use of a similarity index in which a node storing an attribute value related to a similarity of data comprising a first image of a first physical object and a second image of a second physical object is hierarchically configured in such a way as to represent a similarity relation between the data:
• generating, for each piece of the data, a label based on a path indicating a position of the node storing the attribute value in the similarity index;
• storing, for each piece of the data, information associating the attribute value with the label in data storage;
• and
• reconfiguring, for data indicated by whole or part of the information stored in the data storage, the similarity index, based on the label, and executing a similarity-based calculation targeted for the whole or part of the data by use of the reconfigured similarity index,
• wherein storing, distributedly in a plurality of blocks, the information associating the attribute value with the label,
• redistributing the information associating the attribute value with the label in a plurality of blocks, based on a prefix of the label, and then storing the information, and, for data indicated by the information stored in each of redistributed blocks, executing the similarity-based calculation by use of the similarity index which is reconfigured, based on the label, to determine whether the first physical object appearing in the first image is a same physical object as the second physical object appearing in the second image,
• wherein the method improves efficiency in object recognition as to whether the first physical object within the first image is the same object as the second physical object within the second image via reconfiguration of the similarity index and execution of the similarity-based calculation target by use of the reconfigured similarity index.

A non-transitory storage medium storing a program which causes a computer device to execute,

• by use of a similarity index in which a node storing an attribute value related to a similarity of data comprising a first image of a first physical object and a second image of a second physical object is hierarchically configured in such a way as to represent a similarity relation between the data:
• generating, for each piece of the data, a label based on a path indicating a position of the node storing the attribute value in the similarity index;
• storing, for each piece of the data, information associating the attribute value with the label in data storage;
• and
• for data indicated by whole or part of the information stored in the data storage, reconfiguring the similarity index, based on the label, and executing a similarity-based calculation targeted for the whole or part of the data by use of the reconfigured similarity index,
• wherein storing, distributedly in a plurality of blocks, the information associating the attribute value with the label,
• redistributing the information associating the attribute value with the label in a plurality of blocks, based on a prefix of the label, and then storing the information, and, for data indicated by the information stored in each of redistributed blocks, executing the similarity-based calculation by use of the similarity index which is reconfigured, based on the label, to determine whether the first physical object appearing in the first image is a same physical object as the second physical object appearing in the second image,
• wherein the medium improves efficiency in object recognition as to whether the first physical object within the first image is the same object as the second physical object within the second image via reconfiguration of the similarity index and execution of the similarity-based calculation target by use of the reconfigured similarity index.