Face detection method based on context reasoning in unconstrained scene

Abstract

The invention provides a face detection scheme based on context inference in an unconstrained scene, and belongs to the field of multimedia signal processing. According to the invention, data augmentation is carried out on a training set; vGGNet-16 is used as a basic feature extraction network; features of different layers are fused in a weighted mode through a low-level feature pyramid network, acontext auxiliary prediction module is adopted in the prediction link to expand a sub-network so as to deepen and widen a network model, a self-adaptive anchor point sampling data enhancement mode and a multi-scale training method are introduced, and the adaptability of the model to the scale is enhanced. According to the method, description information with the highest expressive force can be extracted, facial features which are not fully extracted can be well made up, the utilization rate of the facial features can be optimized, and the method is suitable for an unconstrained scene with high detection difficulty, and particularly can achieve accurate detection of tiny, fuzzy and shielded faces.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a face detection method based on context reasoning in an unconstrained scene. Background technique [0002] The popularity of intelligent terminal equipment has profoundly affected the way of thinking of human beings, and has a new definition of its social nature. Face detection is the most suitable application for daily life in the field of computer vision. It frees humans from heavy visual processing work and uses machines to analyze and summarize specified information in images and videos. had a profound impact. On smartphones, iPhone X and Huawei Mate20pro realize 3D face recognition unlocking on IOS platform and Android platform respectively, which better protects privacy; in security monitoring, face recognition technology can be used to track and capture criminals and strengthen In terms of property security, Alipay took the lead in launching facial recognition pa...

AI Technical Summary

Problems solved by technology

However, unconstrained human faces in natural scenes are easily disturbed by external environmental factors such as occlusion, illumination, expression, posture, etc., resulting in insufficient extraction and utilization of facial features; The bottleneck lies in the dense sampling of small faces with small anchor points, which can easily generate too many background negative samples, resulting in an increase in false detection rate

The accuracy rate of existing face detection methods in unconstrained scenes is still insufficient, and satisfactory results have not been achieved

Images