[1] Ejaz Ahmed, Michael Jones, Tim K.Marks. An Improved Deep Learning Architecture for Person Re-Identification. In CVPR 2015.
使用dnn来做行人再识别，引入了两个新的layer—Cross-Input Neighborhood Differences layer、Across Patch Features layer。前者用于比较输入的两幅feature map。并按定义的公式算出两种“differences”，后者用于把两种differences整合起来，变成更高阶的，用于衡量两幅图相似度的特征图。Anyway, see the picture of the architecture below !
原文：http://www.cnblogs.com/zhangzph/p/4655621.html