Image recognition system

An image recognition system that recognizes a shape included in an image, the image recognition system comprising:

• an image recognition unit that extracts a feature from an input image based on a result of performing machine learning, and recognizes a target shape included in the input image by using the feature;
• and
• an erroneous recognition determination unit that determines correctness of a recognition result by the image recognition unit,
• wherein the erroneous recognition determination unit comprises:
• a feature importance calculation unit that calculates an importance of the feature;
• a statistical information database that stores a statistic related to the importance;
• and
• a comparison unit that determines the correctness of the recognition result by comparing the importance with the statistic,
• the feature importance calculation unit calculates the importance for each target shape recognized by the image recognition unit and for each type of the feature, by
• using a feature importance parameter representing a magnitude of an influence of the feature when the image recognition unit recognizes the target shape, and
• using a region importance parameter representing a magnitude of an influence of an image region in the input image when the image recognition unit recognizes the target shape, in addition to the feature importance parameter,
• the statistical information database stores the statistic for each type of the target shape and for each type of the feature, and
• the comparison unit determines the correctness of the recognition result for each target shape recognized by the image recognition unit, by comparing, for each target shape recognized by the image recognition unit, the importance calculated by the feature importance calculation unit for each type of the feature with the statistic stored in the statistical information database for each type of feature.