Author Affiliations
School of Electronics and Information, Xi'an Polytechnic University, Xi'an, Shaanxi 710600, Chinashow less
Fig. 1. Person re-identification in video surveillance
Fig. 2. Complete flow chart of person re-identification
Fig. 3. Person's appearance differences. (a) Pose; (b) resolution; (c) lighting; (d) background
Fig. 4. Non-idealized scenes. (a) Occlusion; (b) person misalignment
Fig. 5. Person re-identification verification model
Fig. 6. Person re-identification classification model
Fig. 7. Person re-identification model based on the contrastive loss
Fig. 8. Person re-identification model based on triplet loss
Fig. 9. Method based on pose estimation models
Fig. 10. AlignedReID
[58] calculates the local distance by finding the shortest path
Fig. 11. Difficulties in partial person re-identificaiton. (a) Spatial misalignment; (b) noise from unshared regions
Fig. 12. Unconditional generative model
Fig. 13. Pose-conditioned generative model
Fig. 14. Intra-domain and inter-domain image styles differences
Fig. 15. Accuracy of 14 typical methods in the Market1501
Name | Time | ID | Number of images | Camera | Label method | Evaluation | Size |
---|
VIPeR[21] | 2007 | 632 | 1264 | 2 | Hand | CMC | 128×48 | PRID2011[22] | 2011 | 934 | 24541 | 2 | Hand | CMC | 128×64 | CUHK01[23] | 2012 | 971 | 3884 | 2 | Hand | CMC | 160×60 | CUHK03[14] | 2014 | 1467 | 13164 | 10(5 pairs) | Hand/DPM | CMC | vary | Market-1501[86] | 2015 | 1501 | 32668 | 6 | Hand/DPM | mAP | 128×64 | DukeMTMC-reID[72] | 2017 | 1812 | 36441 | 8 | Hand | CMC | vary | MSMT17[83] | 2018 | 4101 | 126441 | 15 | Faster RCNN | CMC | vary |
|
Table 1. Person re-identification common datasets
Name | Time | ID | Number of images | Gallery | Probe |
---|
Partial iLIDs[66] | 2011 | 119 | 476 | 3 | 1 | Partial ReID[87] | 2015 | 60 | 600 | 5 | 5 |
|
Table 2. Partial person re-identification common datasets