论文笔记 | FaceNet: A Unified Embedding for Face Recognition and Clustering
来源:互联网 发布:有搭接的网络参数计算 编辑:程序博客网 时间:2024/06/01 07:36
Authors
Florian Schroff Dmitry Kalenichenko James Philbin
Florian Schroff
Abstract
本文提出了FaceNet system,直接从face images 学习到 compact Euclidean 欧几里德 space 从而得到face的相似程度。这样一来 face recognition, verification, clustering 就容易了。该方法使用深度卷积网络直接得到embedding,没有用传统的bottleneck层,训练时用到了online triplet mining method。 效率更高,每脸128bytes。
1 Introduction
verification:is this the same person (thresholding)
recognition: who is this person (K-NN classification)
clustering : find common people among these faces (k-means or agglomerative clustering)
本文直接使用squared L2 距离来判断脸部的similarity。
以往使用bottleneck layer 不直接而且representation size 非常大。本文使用triplet based loss function(参考下面的文献)仅128D。其中有两个matching face thumbnail,一个non-matching face thumbnail
K. Q.Weinberger, J. Blitzer, and L. K. Saul. Distance metriclearning for large margin nearest neighbor classification. InNIPS. MIT Press, 2006. 2, 3
tripletsd 的选择很重要,收到curriculum learning的启发,本文使用在线negative exemplar mining 策略,保证了训练过程中难度的增大
Y. Bengio, J. Louradour, R. Collobert, and J. Weston. Curriculumlearning. In Proc. of ICML, New York, NY, USA,2009. 2
3 Method
3.1 Triplet Loss
通过L2之后得到的embedding是一个hypersphere,其约束方程为:
loss为:
3.2 Triplet selection
如果随便选则
1. 离线选择,每n步使用最近的网络再一个subset中选择所需要的样本;
2. 在线选择,mini-batch中选择
本文选择第二种,其中positive的选择时使用所有的正例,(这样会再开始的时候更加稳定收敛速度稍快)。hard negative的选择是
batch size 是1800左右。
3.3 deep convolutional networks
4 Datasets and evaluation
psame=i,j of the same identity
pdiff= different identities
true accepts :TA
false accept: FA
val:TA/psame
FAR:FA/pdiff
5 experiments
图片质量越好越好,Embedding Dimensionality 越大得到 结果应该越好但是需要更多的训练次数,小的dimensionality有利于嵌入移动设备。随着数据的增多结果会有提高但是增加到一定的程度起作用就会变小。
summery
- Authors
- Abstract
- Introduction
- Method
- 1 Triplet Loss
- 2 Triplet selection
- 3 deep convolutional networks
- Datasets and evaluation
- experiments
- summery
- FaceNet: A Unified Embedding for Face Recognition and Clustering 笔记
- 【深度学习论文笔记】FaceNet: A Unified Embedding for Face Recognition and Clustering
- 论文笔记 | FaceNet: A Unified Embedding for Face Recognition and Clustering
- 【论文笔记】FaceNet: A Unified Embedding for Face Recognition and Clustering
- FaceNet: A Unified Embedding for Face Recognition and Clustering
- FaceNet: A Unified Embedding for Face Recognition and Clustering
- FaceNet: A Unified Embedding for Face Recognition and Clustering
- FaceNet: A Unified Embedding for Face Recognition and Clustering论文解读
- 【论文阅读二】FaceNet:A Unified Embedding for Face Recognition and Clustering
- 论文翻译(1)FaceNet: A Unified Embedding for Face Recognition and Clustering
- FaceNet-A Unified Embedding for Face Recognition and Clustering 论文解读
- 人脸识别“FaceNet: A Unified Embedding for Face Recognition and Clustering”
- 【学习】facenet:a unified embedding for face recognition and clustering_wrapper
- 论文阅读:CVPR 2015 FaceNet: A Unified Embedding for Face Recognition and Clustering
- 【论文笔记】SphereFace: Deep Hypersphere Embedding for Face Recognition
- SphereFace : Deep Hypersphere Embedding for Face Recognition学习笔记
- SphereFace: Deep Hypersphere Embedding for Face Recognition
- 人脸识别--SphereFace: Deep Hypersphere Embedding for Face Recognition
- 剑指offer——数组中只出现一次的数字
- C语言程序设计谭浩强版 一
- uc高级编程之创建和删除目录
- Attempt to invoke interface method 'java.lang.Object[] java.util.Collection.toArray()' on a null obj
- Java 并发编程:线程间的协作(wait/notify/sleep/yield/join)
- 论文笔记 | FaceNet: A Unified Embedding for Face Recognition and Clustering
- 转载--小甲鱼PE详解之区块描述、对齐值以及RVA详解(PE详解06)
- 逆向技能+2 JNI介绍
- OpenJudge 百练 2016 ACM 暑期课练习题 滑雪
- 学习技术,最忌讳的就是眼高手低!
- 崔希凡-javaWeb-笔记day07-day09(2016年7月26日23:17:27)
- 进程虚拟内存
- 自写JQ控件-树状菜单控件[demo下载]
- 夏季集训前序