Research map / Label-efficient and robust learning / Recognition and long-tail
Dual-Edged Homogeneous-Modality Similarity: Towards Visible-Infrared Modality-Incomplete Person Re-Identification with Modality Adaptive Matching
X X Xu, Shuhao Zhan, Wei Liu, Zheng Wang, Kui Jiang, Chia‐Wen Lin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add grayscale conversion only in Stage 1 before RGB training (Gray-RGB vs RGB) (ablation) | MAMT · SYSU-VIMI (0.1), VIMC | mAP: 55.79 -> 60.34 mAP, Rank-1 95.09 -> 96.37 | Improved |
| use grayscale images for all training (Gray vs RGB) (ablation) | MAMT · SYSU-VIMI (0.1), VIMC | mAP: 55.79 -> 34.69 mAP, a decrease of 21.10% | Hurt |
| add visible-specific branch (MAM) (ablation) | MAMT with ViT · SYSU-VIMI (0.1), VIMC | Rank-1/mAP: 94.29 -> 95.33 Rank-1, 56.32 -> 57.73 mAP | Improved |
| add infrared-specific branch (MAM) (ablation) | MAMT with ViT · SYSU-VIMI (0.1), VIMC | Rank-1/mAP: 94.29 -> 94.96 Rank-1, 56.32 -> 56.99 mAP | Improved |
| add both visible and infrared branches (ablation) | MAMT with ViT · SYSU-VIMI (0.1), VIMC | Rank-1/mAP: 94.29 -> 96.37 Rank-1, 56.32 -> 59.39 mAP | Improved |
| add divergence loss L_D on top of both branches (ablation) | MAMT with ViT · SYSU-VIMI (0.1), VIMC | mAP: 59.39 -> 60.34 mAP, Rank-1 maintained at 96.37 | Improved |
| add all MAM modules and L_D (ResNet backbone) (ablation) | MAMT with ResNet · SYSU-VIMI (0.1), VIMC | Rank-1/mAP: 89.20 -> 92.76 Rank-1, 37.35 -> 41.59 mAP | Improved |
| add all MAM modules and L_D (ViT backbone) (ablation) | MAMT with ViT · SYSU-VIMI (0.1), VIMC | Rank-1/mAP: 94.29 -> 96.37 Rank-1, 56.32 -> 60.34 mAP | Improved |
| use one unified model instead of multi-model ensemble of AGW | MAMT · VIMI-ReID | memory and time: 20G -> 12G memory, 8h -> 5h training | Improved |
| use MAMT instead of all-in-one IRM under VIMC | MAMT vs IRM · VIMC | mAP: 16.50 -> 60.34 mAP | Improved |
| MAMT vs second-best PMT under VIMC on SYSU-VIMI (0.5) | MAMT · SYSU-VIMI (0.5), VIMC | Rank-1/mAP: Rank-1 99.31 -> 98.91, mAP 67.24 -> 69.50 | Mixed |
| MAMT vs second-best on RegDB-VIMI (0.5) under VIMC | MAMT · RegDB-VIMI (0.5), VIMC | mAP: 84.46 -> 93.52 mAP | Improved |
Closest papers by meaning
- Modality Unifying Network for Visible-Infrared Person Re-Identification (ICCV 2023 · similarity 0.94)
- Empowering Visible-Infrared Person Re-Identification with Large Foundation Models (NeurIPS 2024 · similarity 0.93)
- Learning Progressive Modality-Shared Transformers for Effective Visible-Infrared Person Re-identification (AAAI 2023 · similarity 0.93)
- FMCNet: Feature-Level Modality Compensation for Visible-Infrared Person Re-Identification (CVPR 2022 · similarity 0.92)
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.90)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
