Research map / Detection and segmentation / CNN features
Spindle Net: Person Re-Identification With Human Body Region Guided Feature Decomposition and Fusion
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 body-region-guided features and competitive tree fusion | Spindle Net · CUHK03 | CMC Top-1 accuracy: 78.4 -> 88.5 | Helped |
| Add body-region-guided features and competitive tree fusion | Spindle Net · Market-1501 | CMC Top-1 accuracy: 65.9 -> 76.9 | Helped |
| Add body-region-guided features and competitive tree fusion | Spindle Net without fine-tuning · SenseReID | CMC Top-1 accuracy: 23.0 -> 34.6 | Helped |
| Pool macro and micro features at FEN-C1 and FEN-C2, respectively, rather than use full-image features only (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 72.1 -> 74.7 | Helped |
| Pool both macro and micro features at FEN-C3 (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 72.1 -> 67.8 | Hurt |
| Pool macro features at FEN-C1 and micro features at FEN-C2 instead of reversing those stages (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 74.0 -> 74.7 | Helped |
| Use tree-structured fusion with element-wise max instead of linear fusion with concatenation (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 72.8 -> 76.3 | Helped |
| Use tree-structured fusion with element-wise max instead of inverse-tree fusion with element-wise max (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 41.9 -> 76.3 | Helped |
| Use element-wise max rather than concatenation in tree-structured fusion (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 67.1 -> 76.3 | Helped |
| Use element-wise max rather than average in tree-structured fusion (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 74.3 -> 76.3 | Helped |
| Fine-tune the entire network after tree-plus-max fusion (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 76.3 -> 76.9 | Helped |
| Use linear fusion with element-wise average rather than concatenation (ablation) | Spindle Net · Market-1501 | CMC Top-1 accuracy: 72.8 -> 62.7 | Hurt |
Closest papers by meaning
- Learning Deep Context-Aware Features Over Body and Latent Parts for Person Re-Identification (CVPR 2017 · similarity 0.90)
- Relation Network for Person Re-Identification (AAAI 2020 · similarity 0.89)
- Person30K: A Dual-Meta Generalization Network for Person Re-Identification (CVPR 2021 · similarity 0.89)
- Interaction-And-Aggregation Network for Person Re-Identification (CVPR 2019 · similarity 0.89)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.87)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · 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.