Research map / Vision-language and multimodal / Cross-modal learning
STMI: Segmentation-Guided Token Modulation with Cross-Modal Hypergraph Interaction for Multi-Modal Object Re-Identification
Xingguo Xu, Zhanyu Liu, Weixiang Zhou, Yuansheng Gao, Junjie Cao, Yuhao Wang, Jixiang Luo, Dell Zhang
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 SAM-mask-guided feature modulation (SFM) (ablation) | multi-modal ReID baseline · RGBNT201 | mAP: 70.3 -> 76.1 | Improved |
| add Semantic Token Reallocation (STR) (ablation) | baseline + SFM · RGBNT201 | mAP: 76.1 -> 78.1 | Improved |
| add Cross-Modal Hypergraph Interaction (CHI) (ablation) | baseline + SFM + STR · RGBNT201 | mAP: 78.1 -> 81.2 | Improved |
| fuse modalities with MLP instead of CHI (ablation) | STMI · RGBNT201 | mAP: 78.1 -> 78.0 | No clear effect |
| fuse modalities with self-attention instead of CHI (ablation) | STMI · RGBNT201 | mAP: 78.1 -> 78.4 | No clear effect |
| use layer-wise SFM parameters shared across heads instead of one shared set (ablation) | SFM · RGBNT201 | mAP: 77.2 -> 81.2 | Improved |
| apply SFM to all layers instead of only early layers (ablation) | SFM · RGBNT201 | mAP: 73.5 -> 81.2 | Improved |
| apply SFM to all layers instead of only late layers (ablation) | SFM · RGBNT201 | mAP: 76.7 -> 81.2 | Improved |
| use head-wise SFM parameters (ablation) | SFM · RGBNT201 | mAP: 81.2 -> 77.1 | Hurt |
| add small random mask perturbation for background tokens (ablation) | SFM · RGBNT201 | mAP | Mixed |
| use more than four learnable query tokens (ablation) | STR · RGBNT201 | mAP | No clear effect |
| STMI vs IDEA on RGBNT100 Rank-1 and RGBNT201 Rank-10 | STMI · RGBNT100 / RGBNT201 | Rank-1 / Rank-10: RGBNT201 R-10 93.3 (IDEA) vs 91.6 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- IDEA — IDEA: Inverted Text with Cooperative Deformable Aggregation for Multi-modal Object Re-Identification (CVPR 2025) · better on 5 of 6 shared results
- MambaPro — MambaPro: Multi-Modal Object Re-identification with Mamba Aggregation and Synergistic Prompt (AAAI 2025) · better on 1 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Multi-Modal Object Re-identification via Sparse Mixture-of-Experts (ICML 2025 · similarity 0.88)
- Magic Tokens: Select Diverse Tokens for Multi-modal Object Re-Identification (CVPR 2024 · similarity 0.87)
- UGG-ReID: Uncertainty-Guided Graph Model for Multi-Modal Object Re-Identification (NeurIPS 2025 · similarity 0.86)
- MDReID: Modality-Decoupled Learning for Any-to-Any Multi-Modal Object Re-Identification (NeurIPS 2025 · similarity 0.86)
- MambaPro: Multi-Modal Object Re-identification with Mamba Aggregation and Synergistic Prompt (AAAI 2025 · similarity 0.85)
- DeMo: Decoupled Feature-Based Mixture of Experts for Multi-Modal Object Re-Identification (AAAI 2025 · similarity 0.85)
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.
