This is the official implementation of our paper: Cross-modality Spatial-temporal Transformer for Video-based Visible-infrared Person Re-identification
HITSZ-VCM Dataset: The HITSZ-VCM dataset can be downloaded from this [1] by submitting a copyright form.
-
--dataset: which dataset "VCM ". -
--lr: initial learning rate. -
--gpu: which gpu to run.
First, you may need to manually define the data path. Then, you need to run the code on 1 GPU of A6000 with 48G memory.
Models in Baidu Netdisk.
Password: pz55
This code does not use re_rank.
| Methods | Infrared - Visible | Visible - Infrared |
|---|---|---|
| R1, mAP | R1, mAP | |
| CAJL [3] | 56.59, 41.49 | 60.13, 42.81 |
| MITML [1] | 63.74, 45.31 | 64.54, 47.69 |
| IBAN [2] | 65.03, 48.77 | 69.58, 50.96 |
| Our | 69.44, 51.16 | 72.64, 53.00 |
[1] Lin X, Li J, Ma Z, et al. Learning modal-invariant and temporal-memory for video-based visible-infrared person re-identification[C]. Computer Vision and Pattern Recognition. 2022: 20973-20982.
[2] Li H, Liu M, Hu Z, et al. Intermediary-guided Bidirectional Spatial-Temporal Aggregation Network for Video-based Visible-Infrared Person Re-Identification[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2023.
[3] Ye M, Ruan W, Du B, et al. Channel augmented joint learning for visible-infrared recognition[C]. International Conference on Computer Vision. 2021: 13567-13576.
The code was developed based on the CAJL [1] and MITML[3].
Thanks for [1], [2], [3] providing visible-infrared reid code base and dataset.