Research map / Architectures and efficiency / Attention and transformers
Spatially-Regularized Entropy for Discriminative Token Merging in Fine-Grained Re-Identification
Shangze Li, Yifan Xu, Jingmiao Liang, Yongfei Zhang, Yuzhuo Ma, Yingbo Qu
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 SES-Assess and HCA-Match to ToMe (ablation) | ToMe baseline · Market-1501 | Rank-1: 94.9 -> 95.4 | Improved |
| add EPW-Fuse on top of SES-Assess and HCA-Match (ablation) | SRE-Merge · Market-1501 | Rank-1: 95.4 -> 95.8 | Improved |
| add HCA-Match on top of SES-Assess and EPW-Fuse (ablation) | SRE-Merge · Market-1501 | Rank-1: 95.5 -> 95.8 | Improved |
| add SES-Assess and EPW-Fuse to ToMe baseline (ablation) | ToMe baseline · Market-1501 | mAP: 88.6 -> 89.1 | Improved |
| reduce delta1 from 1 to 1/40 (ablation) | HCA-Function · Market-1501 | Rank-1: 95.8 -> 94.8 | Hurt |
| reduce delta2 from 1 to 1/20 (ablation) | HCA-Function · Market-1501 | mAP: 88.1 -> 89.2 | Improved |
| reduce delta2 further from 1/20 to 1/40 (ablation) | HCA-Function · Market-1501 | mAP: 89.2 -> 89.0 | Hurt |
| use SRE-Merge instead of ToMe at r=17 | ViT-B TransReID · Occluded-Duke | Rank-1: 64.2 -> 67.1 | Improved |
| use SRE-Merge instead of ToMe at r=17 | ViT-B TransReID · MSMT17 | Rank-1: 83.5 -> 84.8 | Improved |
| use SRE-Merge (10.31 GFLOPs) vs ToMe r=6 (16.69 GFLOPs) | ViT-B TransReID · MSMT17 | Rank-1: 84.9 -> 84.8 | No clear effect |
| use SRE-Merge instead of TransReID without compression | ViT-B TransReID · Market-1501 | Rank-1: 95.6 -> 95.8 | No clear effect |
| use SRE-Merge instead of ToMe | DeiT-T · ImageNet-1k | Top-1 accuracy: 71.10 -> 71.43 | Improved |
Closest papers by meaning
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.85)
- Relation-Aware Global Attention for Person Re-Identification (CVPR 2020 · similarity 0.85)
- SSM-Aware Token-Efficient VMamba via Adaptive Patch Pruning and Merging for Person Re-Identification (CVPR 2026 · similarity 0.85)
- Cheb-GR: Rethinking K-nearest Neighbor Search in Re-ranking for Person Re-identification (CVPR 2025 · similarity 0.85)
- A Pose-Sensitive Embedding for Person Re-Identification With Expanded Cross Neighborhood Re-Ranking (CVPR 2018 · similarity 0.85)
- Relation-Guided Spatial Attention and Temporal Refinement for Video-Based Person Re-Identification (AAAI 2020 · similarity 0.84)
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.
