HDCPD-ReID

Hybrid Dynamic Contrast and Probability Distillation for Unsupervised Person Re-Id (HDCPD)

The official repository for Hybrid Dynamic Contrast and Probability Distillation for Unsupervised Person Re-Id. HDCPD achieves state-of-the-art performances on both unsupervised learning tasks and unsupervised domain adaptation tasks for person re-ID. For now, we could only release the test code and model already trained. We will gradually release our training code as the paper delivered.

Prepare Datasets

We use 4 datasets in our train and test, including Market1501, DukeMTMC-ReID, MSMT17, PersonX. Please unzip the datasets under the diretory like

HDCPD-ReID/data
├── market1501
│   └── Market-1501-v15.09.15
├── msmt17
│   └── MSMT17_V1
├── personx
│   └── PersonX
└── duke
    └── DukeMTMC-reID

Training

We utilize 2 tesla v100 GPU for training

To recover UDA results, run:

sh uda.sh

To recover USL results, run:

sh usl.sh

Evaluation

We utilize 1 tesla v100 GPU for testing.

Download Models

You can download model for market1501, duke and personx from Here with access code estr

Unsupervised Learning

To evaluate the model released, run:

CUDA_VISIBLE_DEVICES=0 \
python test.py -d $DATASET \
  --resume $PATH_OF_MODEL

Addition

To recover "Ours+ClusterContrast" results, see "ours+cc" profiles.

Results

Datasets	mAP(%)	R@1(%)	R@5(%)	R@10(%)
Market1501	81.6	92.6	97.4	98.2
DukeMTMC	69.0	82.9	90.9	93
PersonX	84.1	94.4	98.7	99.5
MSMT17	24.6	50.2	61.4	65.7

Acknowledgements

Thanks to SpCL. It is an excellent USL framework and deeply inspires our work.