Method, device, and computer program product for verifying classification result

A device for verifying a classification result, comprising:

• at least one processor;
• and
• memory coupled to the at least one processor and storing instructions, wherein the instructions, when executed by the at least one processor, cause the device to perform actions comprising:
• obtaining a plurality of clusters for training samples by clustering feature representations of the training samples using labels of the training samples;
• determining, based on the plurality of clusters, a clustering class to which an input image belongs;
• acquiring a classification result of a classification model for the input image;
• and
• verifying the degree of reliability of the classification result for the input image based on the clustering class and the classification result;
• wherein determining the clustering class to which the input image belongs comprises:
• calculating, for each cluster in the plurality of clusters, a mean value of the feature representations of the training samples in that cluster and a variance of the feature representations of the training samples in that cluster;
• and
• determining, based on the mean value and the variance calculated for each cluster in the plurality of clusters, a cluster closest to the input image;
• wherein determining the cluster closest to the input image comprises:
• acquiring a feature representation of the input image;
• determining a difference between the feature representation of the input image and the feature representations of the training samples in each cluster in the plurality of clusters, utilizing the mean value and variance calculated for each cluster in the plurality of clusters;
• and
• determining the cluster closest to the input image based on the determined difference.

A computer program product comprising a non-transitory computer-readable storage medium having computer-executable instructions stored therein, wherein the computer-executable instructions, when executed by a computer, cause the computer to perform actions comprising:

• obtaining a plurality of clusters for training samples by clustering feature representations of the training samples using labels of the training samples;
• determining, based on the plurality of clusters, a clustering class to which an input image belongs;
• acquiring a classification result of a classification model for the input image;
• and
• verifying the degree of reliability of the classification result for the input image based on the clustering class and the classification result;
• wherein determining the clustering class to which the input image belongs comprises:
• calculating, for each cluster in the plurality of clusters, a mean value of the feature representations of the training samples in that cluster and a variance of the feature representations of the training samples in that cluster;
• and
• determining, based on the mean value and the variance calculated for each cluster in the plurality of clusters, a cluster closest to the input image;
• wherein determining the cluster closest to the input image comprises:
• acquiring a feature representation of the input image;
• determining a difference between the feature representation of the input image and the feature representations of the training samples in each cluster in the plurality of clusters, utilizing the mean value and variance calculated for each cluster in the plurality of clusters;
• and
• determining the cluster closest to the input image based on the determined difference.