An open set image classification field self-adaption method based on self-paced learning

By employing self-learning and a multi-criteria cross-domain hybrid module, the problem of distinguishing between common and unknown classes in open set image classification is solved, achieving robustness and accuracy under different conditions, and improving the model's transfer efficiency and classification accuracy.

CN116433981BActive Publication Date: 2026-03-20SOUTHEAST UNIV
2 Cites 0 Cited by

Patent Information

Application Number
CN202310427403.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-04-20
Publication Date
2026-03-20
Estimated Expiration
2043-04-20

AI Technical Summary

Technical Problem

Existing unsupervised domain adaptation methods cannot effectively distinguish between common classes and unknown classes in open set image classification, leading to a decline in model performance, and require empirical threshold tuning during the inference stage.

Method used

A self-stepping learning method is adopted to construct a feature extraction module and a dual multi-class classifier module to align shared class features and separate private class features. A multi-criteria cross-domain hybrid module is combined for self-stepping learning to output the classification results of the target domain image, avoiding negative transfer phenomenon, and learning domain-invariant features in the feature space.

Benefits of technology

It achieves good robustness under different hyperparameters and experimental settings, can accurately distinguish between common classes and unknown classes, covers smooth and non-smooth class distributions, requires no empirical parameter tuning, and improves the model's transfer efficiency and accuracy.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application discloses an open set image classification field self-adaption method based on self-step learning, first, the original image is preprocessed to obtain an image set, then a feature extraction module and a double multi-class classifier module are constructed and trained to align shared class features of source domain images and target domain images and separate target domain private class features, a multi-criteria cross-domain hybrid module is further constructed and trained, cross-domain hybrid images are generated by using the source domain images and the target domain images, and the shared class features are self-learned, and finally, a classification result of the target domain image is output. Compared with the existing open set image classification field self-adaption method, the application covers smooth and non-smooth class distribution, and does not need to empirically adjust the threshold for distinguishing common class images and private class images in the inference stage, so that the model has good robustness under different hyperparameters and experimental settings.
Need to check novelty before this filing date? Find Prior Art