Face Age Estimation Based On Label Distribution

With the rapid development of computer technology,the application of computer vision is becoming more and more widespread.Age information,as an important human biometric feature,has numerous application needs in the field of human-computer interaction and has an important impact on the performance of face recognition systems.In recent years,face age estimation techniques based on deep learning have made some development in the field of biometric recognition,but the performance of face age estimation is still not very satisfactory due to the difficulties in collecting the complete data set,the imbalance of image age distribution in different age groups,and the uneven distribution of label estimation probability.In this paper,to address the above problems,for the research of the face age estimation method based on label distribution,the main work and innovation results are as follows:1.The face age estimation method based on counterfactual label distribution learning is proposed for the problem of unbalanced distribution of data sets.This method uses a generative model for data augmentation and a counterfactual generative model to decouple age and identity information to generate more data and avoid the bias caused by uneven data distribution during training.The network can learn evenly for all ages and improve the accuracy of age estimation of faces.2.To address the accuracy problem of the network model,a face label distribution method based on reinforcement learning is proposed.This method uses reinforcement learning to alleviate the phenomenon of multiple peaks in probability distribution images and to eliminate some noise after data expansion,generating interpretable mixed Gaussian distributions,and using reinforcement learning to model the Markovian decision process of characterizing age to biological age,thus enhancing the robustness of the model and improving the accuracy of the model.3.Face age estimation system.Based on the method proposed in this paper,a face age estimation system based on the Django framework is designed and implemented by analyzing the functional and non-functional requirements and feasibility of the system,taking the software design specification as the criterion,and implementing the training face age estimation model,testing the training model and the age estimation function for verifying the effectiveness and practicality of the method in this paper.