论文部分内容阅读
提出了一种应用于跨年龄人脸识别的联合学习方法,该方法由深度卷积神经网络构建而成,能在特征学习的同时学习到最优的测度函数,从而避免不合适的固定阈值所带来的匹配错误.针对有限的内存、过拟合和计算复杂性高的问题,在模型训练过程中采用了多种新颖和有效的训练策略.实验证实了该联合学习方法的有效性,在公开数据库MORPH-II上的识别正确率达到了93.6%.
This paper presents a joint learning method applied to cross-age face recognition, which is constructed by deep convolutional neural network and can learn the best measure function while learning features to avoid inappropriate fixed threshold Aiming at the problems of limited memory, over-fitting and high computational complexity, a variety of novel and effective training strategies are adopted in the model training process.Experiments verify the effectiveness of the joint learning method in The recognition accuracy of the public database MORPH-II reached 93.6%.