Research map / Label-efficient and robust learning / Recognition and long-tail
Group Re-Identification via Unsupervised Transfer of Sparse Features Encoding
Giuseppe Lisanti, Niki Martinel, Alberto Del Bimbo, Gian Luca Foresti
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 |
|---|---|---|---|
| use average pooling instead of max pooling of cosine residuals (ablation) | PREF (500 atoms, E+V+C) · i-LIDS groups | Rank-1 accuracy: 24.6 -> 31.1 | Improved |
| use cosine instead of Euclidean residual distance (average pooling) (ablation) | PREF (500 atoms, E+V+C) · i-LIDS groups | Rank-1 accuracy: 28.5 -> 31.1 | Improved |
| use cosine instead of L1 distance (average pooling) (ablation) | PREF (500 atoms, E+V+C) · i-LIDS groups | Rank-1 accuracy: 30.1 -> 31.1 | Improved |
| train dictionary on ETHZ, VIPeR and CAVIAR instead of CAVIAR only (cosine, average) (ablation) | PREF (500 atoms) · i-LIDS groups | Rank-1 accuracy: 29.6 -> 31.1 | Improved |
| train dictionary on ETHZ+VIPeR+CAVIAR instead of ETHZ only (cosine, average) (ablation) | PREF (500 atoms) · i-LIDS groups | Rank-1 accuracy: 31.2 -> 31.1 | No clear effect |
| increase PCA components from 20 to 50 (ablation) | PREF (500 atoms) · i-LIDS groups | Rank-1 accuracy: 29.4 -> 31.1 | No clear effect |
| use 500 atoms instead of 300 (ablation) | PREF · i-LIDS groups | Rank-1 accuracy: 29.3 -> 31.1 | Improved |
| use 1000 atoms instead of 500 (ablation) | PREF · OGRE | Rank-1 accuracy: 15.1 -> 12.9 | Hurt |
| replace IFV (128 clusters) with PREF (500 atoms) | encoding · i-LIDS groups | Rank-1 accuracy: 26.1 -> 31.1 | Improved |
| replace VLAD (500) with PREF (500) | encoding · Museum groups | Rank-1 accuracy: 23.0 -> 25.8 | Improved |
| replace IFV (64 clusters) with PREF (500 atoms) | encoding · OGRE | Rank-1 accuracy: 14.6 -> 15.1 | Improved |
| concatenate (x,y) patch position to features (ablation) | PREF (500 atoms) · i-LIDS groups | Rank-1 accuracy: 31.1 -> 21.7 | Hurt |
Closest papers by meaning
- Learning Multi-Attention Context Graph for Group-Based Re-Identification (TPAMI 2020 · similarity 0.90)
- Unsupervised Group Re-identification via Adaptive Clustering-Driven Progressive Learning (AAAI 2024 · similarity 0.88)
- Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification (ICCV 2019 · similarity 0.87)
- Uncertainty Modeling for Group Re-Identification (IJCV 2024 · similarity 0.86)
- Dynamic Feature Pruning and Consolidation for Occluded Person Re-identification (AAAI 2024 · similarity 0.85)
- AG-VPReID: A Challenging Large-Scale Benchmark for Aerial-Ground Video-based Person Re-Identification (CVPR 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.
