Facial image age estimation method based on three-level residual error network

Abstract

The invention discloses a facial image age estimation method based on a three-level residual error network, belongs to the field of data processing technology and aims to increase the facial image age estimation level under a non-limited condition. According to the technical scheme, the method comprises the steps that first, the three-level residual error network is established on the basis of a basic residual error network framework; second, the three-level residual error network is adopted to perform pre-training on an ImageNet dataset to obtain an ImageNet residual error network model; third, fine-tuning training is performed on the obtained ImageNet residual error network model on a facial age dataset under the non-limited condition; and last, the three-level residual error network obtained after fine-tuning training is utilized to perform facial image age estimation. According to the method, the three-level residual error network is adopted to realize facial image age estimation, the learning ability of a DCNN network model is greatly improved, the problems of over-fitting and gradient disappearance in the training process are well solved, and therefore the accuracy of facial image age estimation under the non-limited condition is improved.

Description

technical field [0001] The invention relates to a method for accurately estimating age according to face images under unrestricted conditions, and belongs to the technical field of data processing. Background technique [0002] Face is an extremely rich source of information, and people can obtain a lot of useful information from face images, such as identity, gender, age, and expression. As one of the key information of human face, age plays a fundamental role in people's social interaction, so relying on facial images to realize automatic age estimation is one of the important tasks in the field of artificial intelligence. At present, face age estimation has good application prospects in many intelligent fields such as age-based human-computer interaction, access control, visual surveillance, marketing, and law enforcement. [0003] The main idea of ​​face image age estimation is to extract the main features from the face image, and then use classification or regression m...

AI Technical Summary

Problems solved by technology

[0008] 3. A deeper DCNN network will lead to the problem of gradient disappearance, which seriously hinders the learning process of

Images