LEARNING DEVICE, DENSITY MEASUREMENT DEVICE, LEARNING METHOD, LEARNING PROGRAM, AND DENSITY MEASUREMENT SYSTEM

摘要

PROBLEM TO BE SOLVED: To provide data for performing density calculation with high accuracy and low memory.SOLUTION: A first extraction unit 10 extracts a plurality of first portion images from a learning image. A first calculation unit 11 calculates a feature quantity of the first portion image. A search unit 12 searches for an object included in the first portion image and applies a vector representing a relative position between a first position in the first portion image and each of all of the objects included in the first portion image to the feature quantity as a label. A voting unit 13 generates a voting histogram for each of the plurality of first portion images. A learning unit 14 learns a regression model showing a relation between the feature quantity of the first portion image and a relative position of the object included in the first portion image by dividing the feature quantity corresponding to each of the plurality of first portion images to a plurality of clusters so that variation of a corresponding voting histogram is decreased. A first prediction unit 15 predicts, for each of a plurality of clusters, a representative label from labels applied to the feature quantity belonging to the cluster.