Research map / Label-efficient and robust learning / Recognition and long-tail
Hierarchical Gaussian Descriptors with Application to Person Re-Identification
Tetsu Matsukawa, Takahiro Okabe, Einoshin Suzuki, Yoichi Sato
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 hierarchy (covariance of patch covariances) to covariance embedding (ablation) | Cov · VIPeR (XQDA) | rank-1 accuracy: 30.8 -> 35.5 | Improved |
| add hierarchy to zero-mean Gaussian embedding (ablation) | ZmG · VIPeR (XQDA) | rank-1 accuracy: 36.5 -> 48.9 | Improved |
| add hierarchy to Gaussian embedding (ablation) | Gauss · VIPeR (XQDA) | rank-1 accuracy: 37.8 -> 50.9 | Improved |
| use Gaussian (mean plus covariance) instead of covariance only (ablation) | Cov · VIPeR (XQDA) | rank-1 accuracy: 30.8 -> 37.8 | Improved |
| add I-L2 (intrinsic mean removal) norm normalization (ablation) | ZOZ · CUHK03 (XQDA) | rank-1 accuracy: 41.9 -> 80.9 | Improved |
| add I-L2 norm normalization (ablation) | GOG · CUHK03 (XQDA) | rank-1 accuracy: 7.9 -> 80.2 | Improved |
| use I-L2 instead of E-L2 norm normalization (ablation) | GOG · VIPeR (XQDA) | rank-1 accuracy: 47.5 -> 50.5 | Improved |
| add scale normalization without norm normalization (ablation) | ZOZ · Market-1501 single-query (XQDA) | rank-1 accuracy: 53.0 -> 35.4 | Hurt |
| add scale normalization on top of I-L2 (ablation) | ZOZ · VIPeR (XQDA) | rank-1 accuracy: 46.2 -> 48.9 | Improved |
| add scale normalization on top of I-L2 (ablation) | GOG · VIPeR / GRID (XQDA) | rank-1 accuracy: 50.5 -> 50.9 (VIPeR), 28.6 -> 28.3 (GRID) | Mixed |
| add scale normalization with no norm normalization (ablation) | GOG · VIPeR (XQDA) | rank-1 accuracy: 45.9 -> 47.9 | Improved |
Closest papers by meaning
- Hierarchical Gaussian Descriptor for Person Re-Identification (CVPR 2016 · similarity 0.97)
- Learning Instance-Level Spatial-Temporal Patterns for Person Re-Identification (ICCV 2021 · similarity 0.85)
- Semi-Supervised Bayesian Attribute Learning for Person Re-Identification (AAAI 2018 · similarity 0.85)
- Unsupervised Learning of Multi-Level Descriptors for Person Re-Identification (AAAI 2017 · similarity 0.85)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.85)
- Learning Commonality, Divergence and Variety for Unsupervised Visible-Infrared Person Re-identification (NeurIPS 2024 · 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.
