Research map / Architectures and efficiency / Pruning and efficient tuning
SSM-Aware Token-Efficient VMamba via Adaptive Patch Pruning and Merging for Person Re-Identification
Huiyuan Huang, Sang Min Yoon
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 Delta-guided token pruning to VMamba (ablation) | VMamba-Base · Market-1501 | FLOPs (G) / Rank-1: 15.46 -> 3.83 FLOPs; Rank-1 98.2 -> 96.7 | Mixed |
| add token merging only (no pruning) (ablation) | VMamba-Base · Market-1501 | Rank-1 / mAP: 98.2 -> 98.5 Rank-1; 91.6 -> 92.1 mAP | Improved |
| add token merging only (no pruning) (ablation) | VMamba-Tiny · Market-1501 | mAP: 91.6 -> 91.1 | No clear effect |
| add merging on top of pruning (ablation) | VMamba-Base with pruning · Market-1501 | Rank-1 / FLOPs: 96.7 -> 96.7 Rank-1; 3.83 -> 3.82 FLOPs | No clear effect |
| apply pruning on smaller backbone (Tiny) instead of Base | TE-VMamba · Market-1501 | Rank-1: -4.6 Rank-1 (Tiny) vs -1.5 (Base) | Hurt |
| order pruning before merging instead of merging before pruning (ablation) | TE-VMamba-Base · Market-1501 | FLOPs (G): 7.69 -> 3.82 | Improved |
| order pruning before merging instead of applying both simultaneously (ablation) | TE-VMamba-Base · Market-1501 | Rank-1: 95.5 -> 96.7 | Improved |
| use dot-product similarity instead of L2 distance for merging (ablation) | TE-VMamba-Base · Market-1501 | Rank-1: 95.5 -> 96.7 | Improved |
| use cosine or kNN or RBF similarity instead of dot-product (ablation) | TE-VMamba-Base · Market-1501 | mAP: 87.1 -> 87.0 (cosine) | No clear effect |
| apply token adaptation at inference to query and gallery (ablation) | TE-VMamba-Base · CUHK03-NP (Detected) | Rank-1 / FLOPs (G): 94.6 -> 95.4 Rank-1; 15.46 -> 3.82 FLOPs | Improved |
| apply token adaptation at inference to query and gallery (ablation) | TE-VMamba-Base · Market-1501 | Rank-1 / throughput: 98.2 -> 96.7 Rank-1; 322.42 -> 590.07 im/s | Mixed |
Closest papers by meaning
- Distilled Person Re-Identification: Towards a More Scalable System (CVPR 2019 · similarity 0.86)
- Generalizable Person Re-identification via Self-Supervised Batch Norm Test-Time Adaption (AAAI 2022 · similarity 0.86)
- A Novel Unsupervised Camera-Aware Domain Adaptation Framework for Person Re-Identification (ICCV 2019 · similarity 0.86)
- Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification (ICCV 2019 · similarity 0.86)
- Meta Batch-Instance Normalization for Generalizable Person Re-Identification (CVPR 2021 · similarity 0.86)
- Person Re-identification via Recurrent Feature Aggregation (ECCV 2016 · similarity 0.86)
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.
