Large-scale image online clustering system and method based on comparison learning

Abstract

The invention discloses a large-scale image online clustering system and method based on comparison learning. The system comprises an augmentation subsystem, a feature extraction subsystem, an instance level comparison head subsystem and a category level comparison head subsystem. The method comprises the following steps: S1, performing augmentation operation on an original image sample set to obtain two groups of augmentation image sets; s2, constructing a total loss function, taking two groups of augmented image sets as a training set, and training a large-scale online clustering system by adopting a gradient descent optimization method; s3, performing clustering processing on a to-be-processed image sample set by adopting the trained large-scale online clustering system, and taking a category corresponding to the maximum probability output by the category level comparison head subsystem as a clustering result of each image sample; the invention solves the problems that an existing method cannot achieve large-scale online clustering, the two stages of feature extraction and data clustering are not closely related, and error accumulation is likely to occur.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a large-scale image online clustering system and method based on contrastive learning. Background technique [0002] Clustering is a basic unsupervised machine learning method. Its basic idea is to automatically divide the data into several categories according to the characteristics of the data itself and the similarity between the data without relying on external labels, so that each category The data of different types have the same characteristics, and there are obvious differences between the data of different types. Clustering has a wide range of applications in real life: for example, through the analysis of user consumption behavior, users are divided into several groups with different preferences (such as gourmet, technology lovers, etc.), helping merchants to make targeted recommendations and marketing; through Analyze a large number of photos, divide unlabele...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above-mentioned deficiencies in the prior art, the present invention provides a large-scale image online clustering system and method based on contrastive learning, which solves the problem that existing methods cannot realize large-scale online clustering, and the two problems of feature extraction and data clustering The links between the two stages are not close, and the problem of error accumulation is prone to occur

Images