ECCV A Discriminative Feature Learning Approach for Deep Face Recognition
来源:互联网 发布:手机性价比 知乎 编辑:程序博客网 时间:2024/05/22 15:54
A Discriminative Feature Learning Approach for Deep Face Recognition
主要思想:
是一种metric learning的思想,本文提出在原有的网络结构上加入center loss,仅改变原有的损失层,使得网络容易训练和优化网络。本文同时使用softmax loss and center loss两种监督信号,把softmax loss看作是可训练的Separable Features,Center loss可看做能够训练出Discriminative Features. Center loss使得类内聚合,类间松散。
实例:
通过使用在MNIST数据上验证Center loss的作用,在FC上使用2个神经元,使得特征可视化。在LeNets上进行修改,修改的具体参数为
损失函数:
softmax loss function
Center loss
表示类yi的类中心。
问题:在这个公式中,可以看出,要将所有的训练样本导入网络中,才能确定样本类中心。
解决:1、在每次更新计算样本中心时使用mini-batch 2、为了避免由于少量样本的错误标签导致样本中心的扰乱,使用标量a来控制center的学习率。
改变后:
训练算法
实验结果分析:
2维可视化(MNIST数据集)
仅使用softmax loss function
加入 center loss后
通过两个图对边可以看出
1 0
- ECCV A Discriminative Feature Learning Approach for Deep Face Recognition
- A Discriminative Feature Learning Approach for Deep Face Recognition, ECCV16.
- A Discriminative feature learning approach for deep face recognition
- A Discriminative Feature Learning Approach for Deep Face Recognition
- 阅读A Discriminative Feature Learning Approach for Deep Face Recognition
- A Discriminative Feature Learning Approach for Deep Face Recognition
- 人脸识别 - A Discriminative Feature Learning Approach for Deep Face Recognition
- 人脸识别 - A Discriminative Feature Learning Approach for Deep Face Recognition
- A Discriminative Feature Learning Approach for Deep Face Recognition 原理及在caffe实验复现
- 论文阅读:A Discriminative Feature Learning Approach for Deep Face Recognition
- A Discriminative Feature Learning Approach for Deep Face Recognition 的源码部分分析
- A Discriminative Feature Learning Approach for Deep Face Recognition 原理及在caffe实验复现
- A Discriminative Feature Learning Approach for Deep Face Recognition 原理及在caffe实验复现
- centerloss系列文章笔记——人脸识别:A Discriminative Feature Learning Approach forDeep Face Recognition
- Discriminative Deep Metric Learning for Face Verification in the Wild(文献泛读)
- 泛读:CVPR2014:Discriminative Deep Metric Learning for Face Verification in theWild
- DeCAF: A deep convolutional activation feature for generic visual recognition
- A New General Deep Learning Approach for Natural Language Processing
- HTTP返回码中301与302的区别
- Python菜鸟--字符串
- VS2013编写和调用动态链接库
- 查看LINUX进程内存占用情况 top pmap ps
- Mac升级到macOS Sierra 10.12后SVN管理软件Conerstone破解版安装后提示“已损坏”问题
- ECCV A Discriminative Feature Learning Approach for Deep Face Recognition
- STM32内存管理以及堆和栈的理解
- 问题记录
- Windows学习总结(7)——学会CMD命令提示符的重要性
- 服务器的多域名配置
- iOS JSONKit配置
- 多线程的创建和使用
- CryptoJS aes加密,需key 和偏移量 iv
- iOS开发中的这些权限,你搞懂了吗?