Image classification method based on adversarial fusion multi-source transfer learning

Abstract

The invention discloses an image classification method based on adversarial fusion multi-source transfer learning, and mainly solves the problem of low image classification accuracy in the prior art.According to the implementation scheme, the method comprises: 1) establishing a feature extraction network, and extracting image features from an original image file; 2) inputting the image features into a specific domain discriminator and a classifier, and calculating to obtain domain discrimination loss, a pseudo mark of target domain data and classification loss of source domain data; 3) calculating to obtain the sum of MMD distances of all classes in the source domain and the target domain by utilizing the target domain sample pseudo-mark and the source domain sample mark; 4) training thefeature extraction network, the domain discriminator and the classifier by using the sum of the domain discrimination loss, the classification loss and the MMD distance; and 5) sequentially inputtingthe sample to be detected into the trained feature extraction network, the domain discriminator and the classifier, and outputting the category marker of the sample to be detected. The method can effectively improve the classification accuracy of various images, and can be used for image classification under the condition of missing training data markers.

Description

technical field [0001] The invention belongs to the field of image recognition, and in particular relates to an image classification method, which can be used for image classification under absence of training data marks. Background technique [0002] Transfer learning is to "transfer" the knowledge and experience learned in one field to another different but related field to improve the learning efficiency of the model without having to start learning again. Generally, the domain to be classified or predicted is called the "target domain"; the auxiliary domain with a large amount of labeled data is called the "source domain", and there are domain differences between the two. Using transfer learning to study image classification problems has achieved remarkable results at home and abroad. Existing transfer learning methods can be classified into sample-based, feature-based and model-based methods. [0003] Inspired by the two-person zero-sum game in game theory, some schol...

AI Technical Summary

Problems solved by technology

[0007] Although the above method can realize the image classification work under the absence of target domain data labels, it uses the same network to extract features from the source domain data, resulting in the loss of some effective features of the source domain data, which affects the final classification effect

Images