基于多分支CNN和多尺度特征融合的非受控人脸年龄估计

现有的人脸年龄估计不能很好地兼顾全局-局部细节的特征表达，因而非受控人脸年龄估计的精度存在一定的提升空间。为解决此问题，提出了一种基于多分支卷积神经网络(convolutional neural networks,CNN)和多尺度特征融合的非受控人脸年龄估计方法。该方法根据人脸关键点对人脸图片剪裁得到包含人脸的全局图像和分别包含眼睛、鼻子、嘴巴的局部图像;使用多分支CNN网络提取对应的深层全局特征和局部特征，使用多尺度特征融合网络探索局部特征间的相关性信息从而进行局部特征选择;将融合的局部特征与全局特征拼接得到兼顾全局-局部细节的年龄特征;使用softmax损失函数优化模型进行人脸年龄估计。根据MORPH Album2、FG-NET、LAP2016人脸年龄数据集上的实验结果表明，提出的方法是有效的。

Uncontrolled facial age estimation based on multi-branch CNN and multi-scale feature fusion

The existing facial age estimation methods cannot perform well in the feature expression of global and local details, so there is a certain room for improvement in the accuracy of uncontrolled facial age estimation. To solve this problem, this paper proposes an uncontrolled facial age estimation method based on multi-branch convolutional neural networks (CNN) and multi-scale feature fusion. First, the method cuts the face image to obtain a global image containing the face and partial images containing eyes, mouth, and nose, respectively. Second, the multi-branch CNNs are used to extract the corresponding features. The multi-scale feature fusion network is used to explore the correlation information of the local areas for local features selection. Then, the local features are spliced with the global features to obtain an age feature representation that takes into account both global and local details. Finally, the Softmax loss function is used to optimize the model. The experimental results on three datasets (MORPH Album2, FG-NET, and LAP2016) illustrate that the proposed method is effective.