| CPC G06V 10/7715 (2022.01) [G06V 10/22 (2022.01); G06V 10/762 (2022.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01)] | 12 Claims |
|
1. A system comprising circuitry configured to execute a method comprising:
calculating, when an input image is input to a (Convolutional Neural Network (CNN), each of feature maps output individually from a plurality of filters used in specified layers of the CNN, wherein the plurality of filters detects features of the input image;
classifying each of the feature maps calculated individually for the plurality of filters into any one or more of a plurality of clusters based on a value of the calculated each of the feature maps;
calculating, for each of the feature maps, a weight representing a degree of association with a result of identification for the input image by the CNN;
outputting each of transposed feature maps obtained by transposing, for each of the clusters, the feature maps in the cluster based on a result of the classification of each of the feature maps for the input image classified into any of the clusters and on the weight calculated for each of the feature maps;
retrieving, using the transposed feature maps in each of the clusters each of the clusters including the feature maps linked to the plurality of filters as those linked to the transposed feature maps for the input image;
selecting each of the feature maps belonging to each of the retrieved clusters and one or more stored images linked to the each of the retrieved clusters;
generating, for the each of the clusters for the input image, an input image reaction region map based on the transposed feature maps and the weights each for the input image; and
generating, for the each of the retrieved clusters, stored image reaction region maps for the one or more selected stored images from the feature maps belonging to the clusters linked to the stored images.
|