行人检索 Learning a Discriminative Null Space for Person Re-identification
来源:互联网 发布:淘宝精选怎么加入 编辑:程序博客网 时间:2024/05/16 19:49
CVPR 2016
code: http://www.eecs.qmul.ac.uk/~lz/
当前行人检索问题主要关注于 learning the optimal distance metrics,减少同类距离,扩大异类距离。一般来说一个行人的外表用几千维向量来表示其特征,同一个人一般只有几百个样本(主要是以为收集样本比较困难),所以存在 the classic small sample size (SSS) problem。 因为使用小样本集,所以the within-class scatter matrix 变为奇异点,为了避免,采用降维方法,但是这使得学习到得distance metric 不是最优的。
With a small sample size, the within-class scatter matrix becomes singular [4]; to avoid it, unsupervised dimensionality reduction or regularisation are required. This in turn makes the learned distance metric sub-optimal and less discriminative
本文针对 the classic small sample size (SSS) problem ,将同一个人映射为 null space 的一个点。
2 Related Work
行人检索当前主要算法可以分为三大类:
1)设计好的特征, invariant and discriminant features,特征的维度越来越高,几千到几万维
2)学习好的 距离度量 learning robust and discriminative distance metrics or sub-
spaces for matching people across views
3) 深度学习方法,因为缺少大样本,效果不太好
3 Methodology
3.1. Problem Definition
给定一组训练样本,每个样本属于C个人中的一个,将每个样本提取一组向量作为该样本的特征,将其作为一列 x_i(该列的维度比较大),构成一个数据表示矩阵X,这里我们希望学习一个投影矩阵W,将高纬向量 x_i 投影到低纬向量 y_i
The objective of learning a discriminative null space is to learn a projection matrix W. Person re-id can then be performed by computing the Euclidean distance between two projected vectors in the learned discriminative null space
结果对比:
- 行人检索 Learning a Discriminative Null Space for Person Re-identification
- 行人检索“An Improved Deep Learning Architecture for Person Re-Identification”
- 行人检索"Deep Ranking for Person Re-identification via Joint Representation Learning"
- 行人检索 - Embedding Deep Metric for Person Re-identification
- 行人检索--Beyond triplet loss: a deep quadruplet network for person re-identification
- 【行人识别】Deep Transfer Learning for Person Re-identification
- 【行人识别】Deep Transfer Learning for Person Re-identification
- 【行人识别】Deep Transfer Learning for Person Re-identification
- 行人检索 Partial Person Re-identification ICCV2015
- 【Person Re-ID】Margin Sample Mining Loss: A Deep Learning Based Method for Person Re-identification
- Deep Metric Learning for Person Re-Identification
- 行人检索 - Top-push Video-based Person Re-identification
- 【论文笔记】Margin Sample Mining Loss: A Deep Learning Based Method for Person Re-identification
- Learning Mid-level Filters for Person Re-identification
- Deep Ranking for Person Re-identification via Joint Representation Learning
- A Multi-task Deep Network for Person Re-identification
- 阅读A Discriminatively Learned CNN Embedding for Person Re-identification
- 【Person Re-id】A Discriminatively Learned CNN Embedding for Person Re-identification
- duilib 尝试解决透明窗体WebBroswer控件无法显示的问题
- Spring与Struts区别联系
- java网络socket编程(八)之java中BIO实现聊天系统的群聊功能
- Android沉浸式状态栏完全解析
- 苹果手机,电话号码显示为拨号超链接
- 行人检索 Learning a Discriminative Null Space for Person Re-identification
- 存储访问框架
- Android 开发最佳实践 android-best-practices
- poj 1269 判直线共线,平行,求交点
- 反射
- phpstorm10.0.3破解版汉化方法
- 窗口置顶
- SSH系列(一):Spring + Hibernate + SpringMVC 整合
- 原生javascript实现图片自动轮播和点击轮播代码