Progressive Cross-modal Association Learning for Unsupervised Visible-Infrared Person Re-Identification

Paper

Dataset

Put SYSU-MM01 and RegDB dataset into data/sysu and data/regdb, run prepare_sysu.py and prepare_regdb.py to convert the dataset format.

Prerequisites

• Linux
• A minimum of 24GB of GPU memory (Batchsize = 64)

Getting Started

git clone https://github.com/YimingYang23/PCA_USLVIReID.git

install dependencies

• conda create -n PCAReID python==3.7
• conda activate PCAReID
• pip install -r requirement.txt

Training

sh ./run_train_sysu.sh   # for SYSU-MM01
sh ./run_train_regdb.sh  # for RegDB

Testing

Before testing, you need to modify the path to the weights and the path to the dataset in test_sysu.py and test_regdb.py. Additionally, in test_regdb.py, you also need to modify the Trial accordingly.

python ./test_sysu.py    # for SYSU-MM01
python ./test_regdb.py   # for RegDB

Performance

We conducted multiple tests of our method to achieve the best results.

Performance on SYSU-MM01 (Batchsize = 64)

All Search					Indoor Search
Rank-1	Rank-10	Rank-20	mAP	mINP	Rank-1	Rank-10	Rank-20	mAP	mINP
54.39	89.09	95.62	51.95	38.09	59.69	93.90	98.14	66.72	62.44

Performance on RegDB (10 trials, Batchsize = 64)

Trial	Visible to Infrared					Infrared to Visible
	Rank-1	Rank-10	Rank-20	mAP	mINP	Rank-1	Rank-10	Rank-20	mAP	mINP
1	88.83	95.97	97.67	84.86	75.42	86.99	96.26	97.86	82.21	70.99
2	86.41	94.03	95.83	82.95	72.97	88.25	95.10	96.26	83.02	70.57
3	84.76	93.36	95.63	80.81	71.10	83.93	92.67	94.51	79.36	66.78
4	85.92	93.88	96.36	82.32	72.85	86.17	94.03	95.68	81.23	69.01
5	85.29	94.22	96.60	81.29	70.52	85.44	93.83	96.26	80.02	66.13
6	87.09	95.39	97.09	83.01	72.08	86.84	94.22	97.04	81.84	69.57
7	85.05	93.06	95.05	81.01	70.32	84.37	93.79	96.07	79.03	65.78
8	86.12	93.11	95.97	82.96	73.16	88.16	93.59	95.73	82.53	70.20
9	86.31	93.01	95.39	81.82	70.72	84.81	93.16	95.53	80.04	67.35
10	88.54	95.00	96.94	84.09	74.23	87.14	93.93	96.21	83.04	70.77
Average	86.43	94.10	96.25	82.51	72.33	86.21	94.05	96.11	81.23	68.71