Research And Implementation Of Age Estimation Algorithm For The Smart Album System

With the improvement of storage capacity of the photographic equipment,and the improvement of performance of the internet photos sharing platform,how to manage the massive photos has become a problem.The Smart album system arises at the historic moment.A smart album system can`t only manage the album according to face recognition,but also need more management modes.The age estimation in this paper is aim to be applied to the picture grouping and album analysis for the smart album system,to increase the management ability and playability of the smart album.Firstly,we studied the current research status of the existing age estimation algorithms at home and abroad,and decided to use CNN to solve the problem.We decided to choose IMDb dataset.In order to solve the problem of insufficient data,we collected a data set of the age of children on the one hand,and on the other hand,we applied the transfer learning method.Then,we applied the face detection system developed by ourselves,and performed data enhancement operations.Then,the selected basic models were trained locally to verify their availability under the experimental environment.Finally,the age estimation algorithm was optimized by the method of ensemble of deep learning models and the method of multitask learning,and then applied to the smart album system.The contributions of this paper are as follows:(1)Models with different network topology were proposed for the ensemble of deep learning models to optimize age estimation algorithm.The experiment showed that the proposed scheme reduces the MAE by 5.4%.(2)A multitask learning algorithm was proposed to optimize the age estimation algorithm.The experiment proved that the MAE was reduced by 22%.(3)Our age estimation algorithm was applied to the system of the smart album.