| 摘要 |
A method and device for generating an image classifier. The method comprises: acquiring a training sample set, the training sample set comprising N image samples, the N image samples falling within K types, N and K being positive integers, and N being greater than K; acquiring a feature vector of each image sample, wherein the feature vector comprises a hidden variable of the image sample; and based on hidden variables of N image samples, training K types of classifiers via a multivariate logistic regression model. By means of a multivariate logistic regression model, K types of classifiers are simultaneously trained in a maximum likelihood manner, that is to say, the use of the multivariate logistic regression model retains a correlation between the K types of classifiers, and compared with the method whereby an LVSM converts a K-type classification problem in the object classification field into multiple two-type problems isolated from each other, a training result is more accurate. |
