Age and gender prediction method, system and device based on face image

Abstract

The invention provides a method, a system and equipment for carrying out age and gender prediction based on a face image. The method comprises the following steps: acquiring a picture; detecting a face position and a feature point position in the picture by using a face detection algorithm, and correcting and intercepting the face according to preset parameters to obtain a standard face picture; predicting the standard face picture by using a pre-trained neural network model, and outputting the age and gender of the face; the neural network model adopts a feature pyramid structure; according to the method, richer multi-level face features from high to low can be extracted. Meanwhile, age values are expressed as endpoint weight vectors, the endpoint weight vectors serve as middle-layer supervision information in the model training stage to train a model together with age value regression, age regression values are directly output in the prediction stage, and the method is superior to anexisting method in speed and precision; according to the method, a single-stage single-model method is used, age and gender prediction is quickly and accurately carried out on the human face in the video or the picture, and the prediction result can be used for various scenes.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a method, system and device for age and gender prediction based on face images. Background technique [0002] With the rapid development of artificial intelligence technology and the continuous improvement of computer hardware level, the application of deep learning in the field of computer vision is increasing. By predicting the age and gender of faces in pictures collected by ordinary cameras, the data can be fully utilized in scenarios such as security, offline retail, and terminal entertainment, bringing many conveniences to people's lives. [0003] At present, the commonly used method is to predict age and gender separately. This method consumes more time and memory. The same face picture needs to be predicted twice, which is inefficient. Even if a single model is used for age and gender prediction The method is also to extract age and gender features separately, ...

AI Technical Summary

Problems solved by technology

[0003] At present, the commonly used method is to predict age and gender separately. This method consumes more time and memory. The same face picture needs to be predicted twice, which is inefficient. Even if a single model is used for age and gender prediction The method is also to extract the age and gender features separately, which also has the disadvantages of low efficiency and repeated operations.

At present, most of the age predictions are age group predictions, which are classification predictions. The age prediction accuracy of this method is low.

Images