D2C: Deep cumulatively and comparatively learning for human age estimation

作者：

Highlights：

• We propose two supervision signals to improve the performance of deep learning model for human age estimation.

• A novel cumulative hidden layer is proposed to alleviate the sample imbalance problem in human age estimation.

• A novel comparative ranking layer is proposed to facilitate the aging feature learning and improve age estimation.

• The proposed deep model outperforms previous methods by a large margin on two of the largest benchmark datasets.

摘要

Highlights•We propose two supervision signals to improve the performance of deep learning model for human age estimation.•A novel cumulative hidden layer is proposed to alleviate the sample imbalance problem in human age estimation.•A novel comparative ranking layer is proposed to facilitate the aging feature learning and improve age estimation.•The proposed deep model outperforms previous methods by a large margin on two of the largest benchmark datasets.

论文关键词：Age estimation,Deep learning,Convolutional neural network

论文评审过程：Received 14 July 2016, Revised 18 December 2016, Accepted 6 January 2017, Available online 6 January 2017, Version of Record 12 March 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.01.007