改进的VGG网络可提升年龄与性别预测准确率

由于深度卷积网络（Convolutional Neural Network，CNN）具有良好特征学习的性质，它得到了研究者们重点关注，并且已被广泛应用。相比较于深度CNN在物体识别与分类等任务上所达到的出色效果，其在年龄预测与人物性别识别任务上的应用还远远不能令人满意。基于公安业务背景，设计了一个深度卷积网络模型，并在证件照和Adience数据集上训练该模型，从而将其应用在人物年龄预测和性别分类上。通过基于Tensorflow的实验表明，提出的深度卷积网络模型，对人物年龄的预测准确率可达到90%以上；性别分类的准确率也达到93%以上。这明显优于现有文献中的结果。

Improved VGG-Net for Increasing Precision of Age and Gender Prediction

Since deep Convolutional Neural Network（CNN） has good properties of feature learning, it has been investigated deeply and applied widely. Compared with the amazing results applied by deep CNN in object recognition and classification, its application in age prediction and gender discrimination is far away from practice. This paper designs a deep CNN and trains the model on identification photos and Adience dataset, so as to apply it in the prediction of people’s age and gender classification. Experiments based on Tensorflow show that this model is able to successfully estimate individuals’ age at an accuracy rate about 90% and gender at 93% respectively. The results are better than others.