TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification
Authors: Xiao Wang, Lekai Liu, Bin Yang, Mang Ye, Zheng Wang, Xin Xu
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments conducted on the publicly available SYSU-MM01 and Reg DB datasets demonstrate the efficacy of the proposed method. Code https://github.com/liulekai123/Token Matcher. 4 Experiments |
| Researcher Affiliation | Academia | 1School of Computer Science and Technology, Wuhan University of Science and Technology 2 School of Computer Science, Wuhan University 3 Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System, Wuhan University of Science and Technology, China EMAIL,EMAIL |
| Pseudocode | Yes | Algorithm 1: Cross-modality labels Matching |
| Open Source Code | Yes | Code https://github.com/liulekai123/Token Matcher |
| Open Datasets | Yes | Extensive experiments conducted on the publicly available SYSU-MM01 and Reg DB datasets demonstrate the efficacy of the proposed method. |
| Dataset Splits | No | We follow widely used protocols (Ye et al. 2021) to assess the two datasets, where mean precision (m AP), cumulative matching characteristic (CMC), and mean Inverse Negative Penalty (m INP) are adopted. No specific dataset split information (percentages, sample counts, or explicit standard split names with citations) is provided within the paper text itself, other than referring to general 'widely used protocols'. |
| Hardware Specification | No | The numerical calculations in this paper have been done on the supercomputing system in the Supercomputing Center of Wuhan University. This statement indicates a computing environment but lacks specific hardware details like GPU/CPU models, memory, or processor types. |
| Software Dependencies | No | No specific software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9, CUDA 11.1) are explicitly mentioned in the paper. The paper refers to using 'DC-Former (Li et al. 2023)' but this is a method, not a software package with a version number for their implementation. |
| Experiment Setup | Yes | The model is trained over 100 epochs. For the first 50 epochs, we use the baseline method for learning. In the subsequent 50 epochs, we incorporate the proposed framework. Additional details on the settings are provided in the supplementary materials. Hyper-parameter β1, β2 and β3 are weighting parameters to trade-off Lscl, Lneighbor and Lmate in Eq.18, as shown in Figure 3. β1 is set to 0.4 for optimal performance. We also observe that the performance is insensitive to β2 and β3, it achieves the best performance on SYSU-MM01 and Reg DB datasets when β2 = 0.5 or 0.1 and β3 = 0.03 or 0.01. The number of class tokens N is set to 4 for SYSU-MM01 and 6 for Reg DB. |