Face detection method and device applied to mobile terminal

Abstract

The invention provides a face detection method and device applied to a mobile terminal, and the method comprises the steps: building and training an ultra-light-weight face detection model; wherein the ultra-lightweight human face detection model comprises a dw convolution layer, a plurality of weight screening modules, a classification regression module and a non-maximum suppression module, and the dw convolution layer is used for extracting a feature map; wherein each weight screening module comprises a channel domain weight screening module and a spatial domain weight screening module whichare respectively used for screening the feature maps in a channel domain and screening the feature maps after channel domain screening in a spatial domain; the classification regression module is used for classifying and regressing the screened feature maps; the non-maximum suppression module is used for obtaining a face detection result according to the output of all the classification regression modules; and after a to-be-detected face image is obtained, inputting the to-be-detected face image into the trained ultra-lightweight face detection model to obtain a face detection result. According to the invention, face detection can be rapidly realized at the mobile terminal, and the precision is high.

Description

technical field [0001] The present invention relates to the technical field of big data, and in particular, to a face detection method and device applied to a mobile terminal. Background technique [0002] Face detection belongs to the field of target detection, that is, to find the positions of all faces in an image, and frame the faces with a rectangular frame to output the position and size of the rectangular frame. The face detection algorithm can be divided into three stages, namely the early traditional algorithm stage, the AdaBoost framework stage, and the era of deep learning algorithms. Among them, deep learning algorithms are far superior to other methods in terms of speed and accuracy. The mainstream face detection algorithms at this stage include MTCNN, Faceness-Net, FaceR-CNN, SSH, etc. The network of these algorithms can only achieve faster speed under the condition of GPU acceleration, and the model is too large to be suitable for mobile terminals such as mo...

AI Technical Summary

Problems solved by technology

The network of these algorithms often can achieve faster speed only under the condition of GPU acceleration, and the large model is not suitable for mobile terminals such as mobile phones, or the accuracy drops seriously after model compression

Images