CASIA OpenIR  > 智能感知与计算研究中心
Adversarially Occluded Samples for Person Re-identification
Huang HJ(黄厚景)1,2,3,4,5; Li DW(李党伟)1,2,3,4,5; Zhang Z(张彰)1,2,3,4,5; Chen XT(陈晓棠)1,2,3,4,5; Huang KQ(黄凯奇)1,2,3,4,5
2018-06-18
Conference NameIEEE Conference on Computer Vision and Pattern Recognition
Conference DateJune 18-22, 2018
Conference PlaceSalt Lake City, Utah, United States
AbstractPerson re-identification (ReID) is the task of retrieving particular persons across different cameras. Despite its great progress in recent years, it is still confronted with challenges like pose variation, occlusion, and similar appearance among different persons. The large gap between training and testing performance with existing models implies the insufficiency of generalization. Considering this fact, we propose to augment the variation of training data by introducing Adversarially Occluded Samples. These special samples are both a) meaningful in that they resemble real-scene occlusions, and b) effective in that they are tough for the original model and thus provide the momentum to jump out of local optimum. We mine these samples based on a trained ReID model and with the help of network visualization techniques. Extensive experiments show that the proposed samples help the model discover new discriminative clues on the body and generalize much better at test time. Our strategy makes significant improvement over strong baselines on three large-scale ReID datasets, Market1501, CUHK03 and DukeMTMC-reID.
KeywordPerson Re-identification Adversarial Occlusion 行人再识别 对抗 遮挡
Indexed ByEI
Language英语
Document Type会议论文
Identifierhttp://ir.ia.ac.cn/handle/173211/22079
Collection智能感知与计算研究中心
Affiliation1.Institute of Automation, Chinese Academy of Sciences
2.University of Chinese Academy of Sciences
3.CAS Center for Excellence in Brain Science and Intelligence Technology
4.National Laboratory of Pattern Recognition
5.Center for Research on Intelligent Perception and Computing
First Author AffilicationInstitute of Automation, Chinese Academy of Sciences
Recommended Citation
GB/T 7714
Huang HJ,Li DW,Zhang Z,et al. Adversarially Occluded Samples for Person Re-identification[C],2018.
Files in This Item: Download All
File Name/Size DocType Version Access License
Huang_Adversarially_(963KB)会议论文 开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Huang HJ(黄厚景)]'s Articles
[Li DW(李党伟)]'s Articles
[Zhang Z(张彰)]'s Articles
Baidu academic
Similar articles in Baidu academic
[Huang HJ(黄厚景)]'s Articles
[Li DW(李党伟)]'s Articles
[Zhang Z(张彰)]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Huang HJ(黄厚景)]'s Articles
[Li DW(李党伟)]'s Articles
[Zhang Z(张彰)]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: Huang_Adversarially_Occluded_Samples_CVPR_2018_paper.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.