Skip to content

Latest commit

 

History

History
182 lines (153 loc) · 11.6 KB

TOTAL.md

File metadata and controls

182 lines (153 loc) · 11.6 KB

2014

1.Person Re-identification by Video Ranking --ECCV2014 [paper]

特征(feature): HOG3D
本文的主要贡献&创新点:开源了新的视频序列数据集iLIDS-VID;能从有噪声的帧序列中选出关键帧;学习一个视频排序函数.
实验结果(results):
                     iLIDS-VID         PRID
        Rank1          28.9            23.3

2015

1.Sparse Re-ID: Block Sparsity for Person Re-identification --CVPR2015 [paper]

特征(feature): Color Histograms Schmid & Gabor Filters
实验结果(results):
                     iLIDS-VID         PRID
        Rank1          24.9            35.1

2.A Spatio-temporal Appearance Representation for Video-based Pedestrian Re-identification --ICCV2015 [paper]

特征(feature): Fiser vector
实验结果(results):
                     iLIDS-VID         PRID
        Rank1          44.3            64.1

2016

1.Deep Recurrent Convolutional Networks for Video-based Person Re-identification: An End-to-End Approach --arxiv2016 [paper]

特征(feature): four CNN(四层卷积神经网络)
实验结果(results):
                     iLIDS-VID         PRID
        Rank1          42.6            49.8  

2.Top-push Video-based Person Re-identification --CVPR 2016 [paper]

特征(feature): HOG3D+color histograms+LBP
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           56.33          56.74      

3.Person Re-identification by Exploiting Spatio-temporal Cues and Multi-view Metric Learning --IEEE SRL 2016 [paper]

特征(feature): LBP
实验结果(results):
                     iLIDS-VID         PRID
        Rank1          69.13           66.78     

4.Person Re-identification via Recurrent Feature Aggregation --ECCV2016 [paper][code]

特征(feature): LBP+HSV+lab color channels
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           49.3           64.1

5.Recurrent Convolutional Network for Video-based Person Re-identification --CVPR2016 [paper] [code]博客解读:[CSDN]

特征(feature): CNN + RNN  (非常经典的结构)
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           58              70      

6.Temporally aligned pooling representation for video-based person re-identification --ICIP2016 [paper]

特征(feature): Local maximal occurrence representation (LOMO)  时间对齐池化表征(与众不同,关注步态周期循环)
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           55             68.6  

2017

1.Video-based Person Re-identification with Accumulative Motion Contex --TCSVT2017 [paper]

特征(feature): CNN + RNN + opticflow 
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           65.3            78            

2.Learning Compact Appearance Representation for Video-based Person Re-identification --arxiv2017 [paper]

特征(feature): Five CNN (5层卷积神经网络)
实验结果(results):
                     iLIDS-VID         PRID
        Rank1           60.4           83.3

3.See the Forest for the Trees: Joint Spatial and Temporal Recurrent Neural Networks for Video-based Person Re-identification --CVPR2017 [paper]

特征(feature): Temporal Attention Model (TAM) + Spatial Recurrent Model (SRM)
实验结果(results):
                     iLIDS-VID         PRID          Mars
        Rank1           55.2           79.4          70.6
        mAP              --             --           50.7

4.Quality Aware Network for Set to Set Recognition --CVPR2017 [paper] [code]

特征(feature): CNN(GoogleNet)
实验结果(results):
                     iLIDS-VID         PRID          
        Rank1           68.0           90.3       

5.Jointly Attentive Spatial-Temporal Pooling Networks for Video-based Person Re-identification --ICCV2017 [paper] [code]博客解读:[CSDN]

特征(feature): (3层)CNN + RNN + attention
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           62              77            44   

6.A Two Stream Siamese Convolutional Neural Network For Person Re-identification --ICCV2017 [paper]

特征(feature): CNN + RNN + attention  => siamese network
实验结果(results):
                     iLIDS-VID         PRID          
        Rank1           60              78       

7.Deep Cross-Modality Alignmeant for Multi-Shot Person Re-Identification --MM2017 [paper]

特征(feature): (3层)CNN + RNN + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           60              80           63     

8.Data Generation for Improving Person Re-identification --MM2017 [paper]

特征(feature): (3层)CNN + RNN + avgpooling
实验结果(results):
                     iLIDS-VID         PRID                
        Rank1           66              79            

9.Three-Stream Convolutional Networks for Video-based Person Re-Identification --arxiv2017 [paper]

特征(feature): (4层)CNN + RNN + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           67.5           79.7          45.6   

10.Phasic Maximal and Local Maximal Occurrence Representation for Video-Based Person Re-identification --ICCSN2017 [paper]

特征(feature): Phasic Maximal and Local Maximal Occurrence (PM-LOMO)
实验结果(results):
                     iLIDS-VID         PRID                 
        Rank1           57            82.58          

2018

1.TRegion-based Quality Estimation Network for Large-scale Person Re-identification --AAAI2018 [paper]

特征(feature): (4层)CNN + RNN + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           76.1           92.4          77.83            
        mAP              --             --           71.14

2.Video Person Re-identification by Temporal Residual Learning --arxiv2018 [paper]

特征(feature):  GoogleNet + BiLSTM
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           57.7           87.8          79.3            

3.Diversity Regularized Spatiotemporal Attention for Video-based Person Re-identification --CVPR2018 [paper] [code]

特征(feature): Resnet50 + attention
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           80.2           93.2          82.3            
        mAP              --             --           65.8

4.Multi-shot Pedestrian Re-identification via Sequential Decision Making --CVPR2018 [paper] [code]

特征(feature): Inception-BN or AlexNet
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           60.2           85.2          71.2            

5.Exploit the Unknown Gradually: One-Shot Video-Based Person Re-Identification by Stepwise Learning --CVPR2018 [paper] [code]

特征(feature): ResNet50 + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars          DukeMTMC-VideoReID      
        Rank1           --              --          62.67               72.79 
        mAP              --             --          42.45               63.23

6.Video Person Re-identification with Competitive Snippet-similarity Aggregation and Co-attentive Snippet Embedding --CVPR2018 [paper] [code]

特征(feature): ResNet50 + attention
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           85.4           93.0          86.3           
        mAP             87.8           94.5          76.1 

7.SCAN: Self-and-Collaborative Attention Network for Video Person Re-identification --CVPR2018 [paper]

特征(feature): ResNet50 + attention
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           88.0           95.3          87.2           
        mAP             89.9           95.8          77.2 

8.Video-based Person Re-identification via 3D Convolutional Networks and Non-local Attention --arxiv2018 [paper]

特征(feature): ResNet50-3D + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1           81.3           91.2          84.3           
        mAP              --             --            77 

9.Spatial-Temporal Synergic Residual Learning for Video Person Re-Identification --arxiv2018 [paper]

特征(feature): CNN + RNN + avgpooling
实验结果(results):
                     iLIDS-VID         PRID          Mars        
        Rank1            70             88           76.7