ICME2014-MULTI-VIEW GAIT RECOGNITION WITH INCOMPLETE TRAINING DATA
来源:互联网 发布:始作俑者其无后乎意思 编辑:程序博客网 时间:2024/05/02 00:55
主要思想:视角转换模型VTM基础上,考虑了如果缺少部分训练样本(比如视角1中存在的个体样本,在视角2中缺少了),提出了视角特征恢复模型VFRM,然后使用已有的VTM去识别多视角步态。
VFRM使用了同一视角下的K近邻样本来恢复,这个K近邻指的是测地距离(使用Dijkstra方法计算距离)
1.Introduction
仍然是worapan提到的3种分类方法实现多视角步态识别。本文属于VTM这一类,指出现有缺点,本文主要特色就是考虑缺失样本情况下的多视角
2.VTM Based GAIT RECOGNITION
2.1Gait Feature Extraction :
GEI+LDA
2.2View transformation model construction
2.3 Gait Similarity Measurement
3.VFRM BASED Gait Recognition with imcomplete training data
3.1 Geodesic distances based K-nearest-neighbor
K近邻,距离越近,权值越重
3.2 Pedestrians neighborhood measurement
距离的计算: 视角相差越小,权值越大。 Dij= sum(sum(w*dij [Vi,Vj]))遍历各个视角,任意两个视角下 i,j的距离。dij非零。
【PS:如果这种情况,任意两个视角下,i样本与j样本不同时存在,则Dij不能计算出来,这个时候怎么办?】
4.EXPERIMENTS
5. CONCLUSIONS
VFRM使用了同一视角下的K近邻样本来恢复,这个K近邻指的是测地距离(使用Dijkstra方法计算距离)
1.Introduction
仍然是worapan提到的3种分类方法实现多视角步态识别。本文属于VTM这一类,指出现有缺点,本文主要特色就是考虑缺失样本情况下的多视角
2.VTM Based GAIT RECOGNITION
2.1Gait Feature Extraction :
GEI+LDA
2.2View transformation model construction
2.3 Gait Similarity Measurement
3.VFRM BASED Gait Recognition with imcomplete training data
3.1 Geodesic distances based K-nearest-neighbor
K近邻,距离越近,权值越重
3.2 Pedestrians neighborhood measurement
距离的计算: 视角相差越小,权值越大。 Dij= sum(sum(w*dij [Vi,Vj]))遍历各个视角,任意两个视角下 i,j的距离。dij非零。
【PS:如果这种情况,任意两个视角下,i样本与j样本不同时存在,则Dij不能计算出来,这个时候怎么办?】
4.EXPERIMENTS
5. CONCLUSIONS
0 0
- ICME2014-MULTI-VIEW GAIT RECOGNITION WITH INCOMPLETE TRAINING DATA
- 2014LuJW-Human Identity and Gender Recognition from Gait Sequences with Arbitrary Walking Directions
- 2012 Multi-University Training Contest 4-1001 hdu4331 Image Recognition
- Pattern Recognition For HandWritten with Semeion Data
- Multi-view Convolutional Neural Networks for 3D Shape Recognition
- Start Training YOLO with Our Own Data
- Training in YOLOv2 with your data
- hdu 4331 Image Recognition(DP+线段树)2012 Multi-University Training Contest 4
- HDU 6093 Rikka with Number (2017 Multi-University Training Contest
- 论文阅读:Multi-view Convolutional Neural Networks for 3D Shape Recognition
- Discretized Continuous Speech Emotion Recognition with Multi-Task Deep Recurrent Neural Network
- MXNet学习5——Data Parallelism with Multi-devices
- 4.3 Populating a Table View with Data
- Funnel-structured cascade for multi-view face detection with alignmentawareness
- M.Goffredo -Self-Calibrating View-Invariant Gait Biometrics读后记
- 2016 Multi-University Training
- Improving Multi-frame Data Association with Sparse Representations for Robust Near-online Multi-ob
- 读后记GAIT RECOGNITION BASDED ON CONTOURLET TRANSFORM AND COLLABORATIVE REPRESENTATION
- 数组-06. 找出不是两个数组共有的元素(20)
- 4.磁盘分区结构划分
- Spring bean定义解析源码分析
- YT14-HDU-S与D的故事
- 1.计算机病毒的定义、特征、程序结构、命名、传播与生命周期
- ICME2014-MULTI-VIEW GAIT RECOGNITION WITH INCOMPLETE TRAINING DATA
- 2.计算机病毒的分类
- Android 事件处理全面剖析
- github使用入门
- poj 2945 Find the Clones
- 3.流氓软件
- 硬盘逻辑锁
- 可重入内核
- 教你手杀顽固病毒,从此不用重装系统