Face detection method and face detection device, storage medium

Abstract

The invention relates to a face detection method and a face detection device, and a storage medium. In the first aspect, a selection mechanism is adopted in the face detection method, and a better processing method is selected from the face recognition processing and the face tracking processing according to the previous detection result, which is beneficial to enhance the practical effect of theface detection method; In the second aspect, the lightweight depth neural network of face recognition is introduced in the process of face recognition, which can effectively recognize and locate the face region and improve the detection accuracy. In the third aspect, face confidence is introduced to solve the drift problem in the face tracking phase, and the tracking bias is corrected to improve the output accuracy of the face region. In the fourth aspect, the ROI prediction method is added on the basis of the lightweight depth neural network, which can avoid the time-consuming situation caused by the face recognition of the whole image, improve the execution speed of face recognition processing and reduce the overhead of the system.

Description

technical field [0001] The invention relates to face detection technology, in particular to a face detection method, a face detection device, and a storage medium. Background technique [0002] With the development of electronic technology, face detection and recognition has become the most potential means of biological identity verification, requiring automatic face recognition systems to have a certain ability to recognize general images, and a series of problems faced by this make face Detection started as an important research topic. At present, face detection is a key link in the automatic face recognition system, and its application background has gone far beyond the scope of the face recognition system, and has important applications in content-based retrieval, digital video processing, video detection, etc. value. [0003] Face detection is a necessary pre-processing step in the fields of face beauty, face special effects, face recognition, face attribute analysis,...

AI Technical Summary

Problems solved by technology

For example, face detection based on the RCNN series abandons the method of generating candidate regions by sliding windows, and uses the proposal method. Although this method can obtain high detection rates, it has the disadvantages of complex composition of deep neural networks and slow detection speed; The face feature extraction and classification in the face detection of the cascaded CNN mode is often completed by CNN. It is necessary to set up 6 CNNs in the cascaded structure, and use 3 CNNs to classify and judge faces and non-faces. It reduces the time-consuming classification and judgment, which is not conducive to the rapid realization of face detection.

Images