Lightweight face detection method and system based on mixed attention feature pyramid structure

Abstract

The invention discloses a lightweight face detection method and system based on a mixed attention feature pyramid structure, and the method comprises the following steps: carrying out the zooming processing and data enhancement of a small-size face training set image, and obtaining an image of which features are to be extracted; extracting picture features by using a lightweight convolutional neural network; sampling the features on a plurality of feature scales and fusing the features of different scales; processing the feature layer by using a residual bottleneck layer structure, and obtaining a final detection prediction layer through a mixed attention pyramid structure; calculating the prediction layer by using an anchor-free detection method and a focusing loss function, and regressing a face position contained in the image; designing and realizing a face detection system. The system uses the algorithm provided by the invention to carry out rapid face picture detection and video detection. The method has robustness for detection of shielded and multi-scale dense faces in a complex scene, and the effectiveness of the method is proved by a test result and a corresponding face detection picture.

Description

technical field [0001] The invention relates to the technical field of face detection based on deep learning, in particular to a face detection method and system based on a mixed attention feature pyramid structure. Background technique [0002] Face detection refers to the process of determining the position, size and pose of all faces in the input image. It is a key technology in face information processing and has become a research hotspot in the field of computer vision. As a specific application of object detection and one of the key steps of face recognition, face detection has become an independent research direction in vision tasks and has received extensive attention. Today, due to the improvement of computer computing power and the improvement of face detection datasets, deep learning has become the mainstream method for solving computer vision tasks and has achieved outstanding results in the field of face detection. At present, face detection based on deep learn...

AI Technical Summary

Problems solved by technology

[0004] The current face detection tasks often need to deal with dozens or hundreds of face targets. These targets are in random real scenes and have the characteristics of multi-scale, high occlusion, and denseness. At present, most face detection methods use large convolutional neural networks. The network extracts image features, which greatly improves the detection accuracy, but it is also accompanied by problems such as large number of parameters, complex training, and slow detection speed, making it difficult to use in real-time detection scenarios

Although there are some detection methods using lightweight networks, only using lightweight networks to improve detection speed greatly reduces detection accuracy, and these fast face detection methods are difficult to deal with large changes in face scale and large numbers of small-sized faces. complex scene

Images