Multi-view depth generation image clustering method

Abstract

The invention discloses a multi-view depth generation image clustering method, which belongs to the technical field of image clustering and data mining, and comprises the following steps of: 1) pre-training an independent encoding and decoding network of each view, and discovering a potential feature space of each view; 2) pre-training a multi-view feature fusion coding and decoding network, and discovering a multi-view fusion feature space; 3) randomly initializing Gaussian mixture model parameters; and 4) calculating the probability that a data sample is generated by a certain sub-Gaussian model as an image clustering result, generating clustering loss, calculating a Gaussian mixture model parameter updating value, and updating parameters until convergence. The invention designs a multi-view depth generation image clustering method for image data, mainly considers the learning of optimizing features by utilizing complementary information in the multi-view image data, improves the image clustering and feature learning effects, and designs a multi-view feature fusion strategy for the multi-view feature clustering method which fuses data information in a plurality of views end to end. The strategy can effectively improve the utilization rate of multi-view data information and improve the performance of an image clustering algorithm.

Description

technical field [0001] The invention belongs to the technical field of image clustering and data mining, and relates to an image clustering method for multi-view depth generation. Background technique [0002] With the rapid development of global informatization, human beings are entering the era of big data. The amount of information data on the Internet is increasing exponentially every day, such as: multimedia data (images, voice, video, etc.) on the Internet, real-time data received by robot terminals, data from IoT device sensors, etc. The amount is large, the generation speed is fast, the data distribution is uneven, and the information quality inside the data is not very high. Facing the challenges brought by big data, it is an urgent need to quickly analyze and find the potential relational structure and semantic features of data from diversely distributed low-quality data. Therefore, it is necessary to conduct in-depth research on the algorithms and models of data...

AI Technical Summary

Problems solved by technology

[0004] However, the above methods only use the information of the image data itself, and do not consider the prior knowledge between the image data, that is, the multi-view characteristics of the data.

Because they do not consider the information of objects reflected by different images in the data in different perspectives, they only apply the clustering method to the single-view feature of the data, and cannot use the complementary information between multiple perspectives to optimize feature learning, making the performance limited to It is within the range supported by the information of a single perspective, which makes it impossible to make full use of the information contained in the multi-view data

Images