A distributed large-scale face clustering method and device

Abstract

The embodiment of the present application discloses a distributed large-scale face clustering method and device. In the technical solution provided by the embodiment of the present application, the face images to be clustered are clustered in batches and the clustering results are summarized to obtain a clustered set and an unclustered set, and settings are extracted from each class of the clustered set A number of face pictures form a representative class, and calculate the similarity distance between each unclustered face picture and each representative class to obtain a similarity distance set, and cluster the unclustered face pictures into In the most similar representative class, and determine the link relationship between each unclustered face picture and the corresponding representative class according to the set link threshold, merge each representative class based on the link relationship, and output the merged result. Using the above technical means can reduce the amount of data calculation for large-scale face data clustering, reduce memory consumption, and improve the efficiency of face clustering on the premise of ensuring the stability of clustering results.

Description

technical field [0001] The embodiments of the present application relate to the technical field of data clustering, and in particular to a distributed large-scale face clustering method and device. Background technique [0002] At present, in many business scenarios of intelligent security, it is necessary to cluster a large amount of unlabeled face data to simplify the face data labeling process or achieve other data processing purposes. The core problem of face clustering is to divide the collected N unlabeled photos into M individuals, and "M" is unknown in advance. Most of the face clustering algorithms obtain face pictures, use the existing feature extraction technology to extract the feature vector of the face picture, and cluster all the face pictures to be clustered according to the extracted feature vector and a certain clustering algorithm . When the clustering algorithm clusters the clustered face pictures, it often needs to compare the face features to obtain t...

AI Technical Summary

Problems solved by technology

When the magnitude of face data is large, the computational complexity of data clustering increases accordingly, and a large amount of storage space is required to store the similarity value of any two face data

Its data clustering takes a relatively long time, and the increase in memory consumption will lead to insufficient system memory, further affecting the efficiency of face data clustering

Images