Face image age estimation method, device and terminal equipment

Abstract

The invention belongs to the technical field of convolutional neural networks, and provides a face image age estimation method, device and terminal equipment. The method includes building a convolutional neural network model including a potential factorization layer; initializing the convolutional neural network model; inputting preprocessed images to the initialized convolutional neural network model, and training the initialized convolutional neural network model through a counterpropagation method based on an age loss function according to the preprocessed images; and inputting a face image to be detected to the trained convolutional neural network model, and outputting the age of the face in the face image to be detected. The potential factorization layer decomposes features of the image into components related with age which need to be obtained and related components unrelated with age, so training and detection can be performed based on the components related with age, so that the convolutional neural network model has relatively good robustness.

Description

A face image age estimation method, device and terminal equipment thereof technical field The invention belongs to the technical field of convolutional neural networks, and in particular relates to a face image age estimation method, device and terminal equipment thereof. Background technique Face age estimation has great application value in security monitoring, human-computer interaction, video retrieval and other fields. However, face age estimation is affected by many factors such as genes, living environment, and health status. Convolutional neural network has become a research hotspot in recent years due to its excellent performance in large-scale image processing. Among the existing face age estimation methods based on convolutional neural networks, there are mainly active appearance model (AAM), feature subspace model (AGES), appearance model (mainly BIF bionic features), and flow model. However, the active appearance model trains the shape and texture features ...

AI Technical Summary

Problems solved by technology

However, the active appearance model trains the shape and texture features separately, losing the information of the common parts of the shape and texture features, and the active appearance model depends on each key point. Once the key points are not detected in place, the final performance will be affected.

The disadvantage of the feature subspace model is that when building the feature subspace model, multiple images of the same person with different ages are required to build the subspace, so the model will not be suitable for data with only a single age image

The disadvantage of the manifold model is that the manifold learning method of age characteristics requires a large number of databases, which is only suitable for large age databases, and the data distribution of each age group is required to be relatively uniform

The appearance model is the most used model. However, the features extracted by the appearance model are only hand-designed features, and the performance of face age estimation is very limited.

The existing face age estimation method based on the convolutional neural network model is poor in robustness, lacks direct modeling of face features and other features, and the learning ability of the network model is limited

Images