A Face Age Estimation Method Based on Convolutional Neural Networks for Metric Learning

Abstract

The invention discloses a face age estimation method based on convolutional neural network for metric learning. The overall steps are: constructing a data set; dividing the data set into a training set and a verification set; performing mini-batch at the network input layer Constructed in pairs, and then sent to two twin networks for training; constructing the VGG-16 network; network training; using softmax loss and revised contrastive loss together as supervisory signals to adjust the network; network evaluation; the final estimated age is The maximum probability obtained by the softmax layer corresponds to the category. The present invention combines deep learning with metric learning. By introducing metric learning, the feature space has a higher degree of discrimination, and the robustness of the age estimation algorithm is stronger; deep learning is used to combine the feature extraction task with the objective function optimization task, Implementing end-to-end training for the entire task can achieve better performance when applied to public datasets.

Description

technical field [0001] The invention relates to an estimation method, in particular to a face age estimation method based on a convolutional neural network for metric learning. Background technique [0002] As an important biological characteristic of human beings, age information has many application requirements in the field of human-computer interaction, and has an important impact on the performance of face recognition systems. Age estimation based on face images refers to the application of computer technology to model the changes in face images with age, so that the machine can infer the approximate age or age range of a person based on the face image. Using deep learning technology, end-to-end feature extraction and objective function optimization is an important method for face age estimation. However, most of the current methods ignore the order of age and the relationship between ages, making it difficult for features to contain these useful information for estima...

AI Technical Summary

Problems solved by technology

The problem with this technique is that the feature extraction and optimization functions are performed separately, so the extraction of features does not serve the optimization function better.

In addition, although the age distribution of this technology is not obtained by assuming variance, the age label of a single adjacent sample still has uncertainty, so its age distribution is still not ideal

Images