LUPerson-NL

Large-Scale Pre-training for Person Re-identification with Noisy Labels (LUPerson-NL)

The repository is for our CVPR2022 paper Large-Scale Pre-training for Person Re-identification with Noisy Labels.

LUPerson-NL Dataset

LUPerson-NL is currently the largest noisy annotated Person Re-identification dataset without humuan labelling efforts, which is used for Pre-training. LUPerson-NL consists of 10M images of over 430K identities extracted from 21K street-view videos and covers a much diverse range of capturing environments.

Details can be found at ./LUP-NL.

Pre-trained Models

Model	link
ResNet50	R50 code:pr50
ResNet101	R101 code:r101
ResNet152	R152 code:r152

Finetuned Results

For MGN with ResNet50:

Dataset	mAP	cmc1	link
MSMT17	68.0	86.0	-
DukeMTMC	84.3	92.0	-
Market1501	91.9	96.6	-
CUHK03-L	80.4	80.9	-

For MGN with ResNet101:

Dataset	mAP	cmc1	path
MSMT17	70.8	87.1	-
DukeMTMC	85.5	92.8	-
Market1501	92.5	96.9	-
CUHK03-L	80.5	81.2	-

For MGN with ResNet152:

Dataset	mAP	cmc1	path
MSMT17	71.6	87.5	-
DukeMTMC	85.6	92.4	-
Market1501	92.7	96.8	-
CUHK03-L	80.6	81.2	-

Citation

If you find this code useful for your research, please cite our paper.

@article{fu2020unsupervised,
  title={Unsupervised Pre-training for Person Re-identification},
  author={Fu, Dengpan and Chen, Dongdong and Bao, Jianmin and Yang, Hao and Yuan, Lu and Zhang, Lei and Li, Houqiang and Chen, Dong},
  journal={Proceedings of the IEEE conference on computer vision and pattern recognition},
  year={2021}
}

@article{fu2022large,
  title={Large-Scale Pre-training for Person Re-identification with Noisy Labels},
  author={Fu, Dengpan and Chen, Dongdong and Yang, Hao and Bao, Jianmin and Yuan, Lu and Zhang, Lei and Li, Houqiang and Wen, Fang and Chen, Dong},
  journal={arXiv preprint arXiv:2203.16533},
  year={2022}
}