ChunfeiMa/AlignedReID

Official Reproduce AlignedReID: Surpassing Human-Level Performance in Person Re-Identification using Pytorch.

AlignedReID

Official Reproduce AlignedReID: Surpassing Human-Level Performance in Person Re-Identification using Pytorch.

@article{zhang2017alignedreid,
  title={Alignedreid: Surpassing human-level performance in person re-identification},
  author={Zhang, Xuan and Luo, Hao and Fan, Xing and Xiang, Weilai and Sun, Yixiao and Xiao, Qiqi and Jiang, Wei and Zhang, Chi and Sun, Jian},
  journal={arXiv preprint arXiv:1711.08184},
  year={2017}
}

Market1501

Model	Param Size (M)	Loss	Distance	Rank-1/ mAP(%)	RK:Rank-1/ mAP (%)
Resnet50	25.05	softmax	Global	81.2/64.2	83.4/76.4
Resnet50	25.05	softmax+label smooth	Global	82.6/64.4	84.0/76.8
Resnet50	25.05	softmax+trihard	Global	86.4/70.9	88.5/83.3
Resnet50	25.05	AlignedReID	Global	87.5/72.5	89.0/84.7
Resnet50	25.05	AlignedReID	Local	87.5/71.9	89.6/84.9
Resnet50	25.05	AlignedReID	Global+Local	88.4/73.2	90.2/85.5
Resnet50	25.05	AlignedReID(Mutual)	Global	88.2/73.1	89.5.2/84.7

Prepare data

Create a directory to store reid datasets under this repo via

cd AlignedReID/
mkdir data/

If you wanna store datasets in another directory, you need to specify --root path_to_your/data when running the training code. Please follow the instructions below to prepare each dataset. After that, you can simply do -d the_dataset when running the training code.

Market1501 :

1. Download dataset to data/ from http://www.liangzheng.org/Project/project_reid.html.
2. Extract dataset and rename to market1501. The data structure would look like:

market1501/
    bounding_box_test/
    bounding_box_train/
    ...

3. Use -d market1501 when running the training code.

Train

python train_class.py  -d market1501 -a resnet50 

python train_alignedreid.py  -d market1501 -a resnet50 --test_distance global_local

Note: You can add your experimental settings for 'args'

Test

python train_alignedreid.py -d market1501 -a resnet50 --evaluate --resume saved-models/best_model.pth.tar --save-dir log/resnet50-market1501 --test_distance global_local (--reranking)

Note: (--reranking) means whether you use 'Re-ranking with k-reciprocal Encoding (CVPR2017)' to boost the performance.