Face detection method based on clustering analysis and model compression

Abstract

The invention relates to a face detection method based on clustering analysis and model compression. The method comprises the steps: firstly obtaining a face detection data set, carrying out K-means clustering analysis of the face data set, carrying out analysis of the number and a size of anchor frames and an application degree of the data set, and finally generating a detection anchor frame mostsuitable for the data set; applying the generated detection anchor frame to a deep learning network, and training a face detection network; finally, after layer pruning or channel pruning is performed on the trained face detection network, carrying out fine adjustment on the network, and acquiring a lighter-weight network; and detecting the image and the video by using the network to obtain a final result. The method has advantages of high detection accuracy and a high detection rate, and can be applied to human face attendance, access control systems, traffic identity approval and other scenes.

Description

technical field [0001] The invention relates to the fields of face detection and computer vision, in particular to a face detection method based on cluster analysis and model compression. Background technique [0002] With the continuous development and progress of science and technology, information technology, which is an important part of it, has also made great progress, and a large amount of data has emerged. With the birth of data, computer vision technology has also played an increasingly important role in people's lives. role. Today, the technology of face detection is used in various aspects, such as attendance, access control, and even the search for specific groups of people under surveillance. However, the face detection performed by many deep learning networks has the problem of insufficient real-time performance. Contents of the invention [0003] In view of this, in view of the insufficient real-time effect of face detection performed by many deep learning...

AI Technical Summary

Problems solved by technology

[0003] In view of this, in view of the insufficient real-time effect of face detection performed by many deep learning networks, the purpose of the present invention is to provide a face detection method based on cluster analysis and model compression, which will be more in line with the human face through clustering. The anchor frame of face size and quantity is applied to the original deep learning network to improve the accuracy

Images