Research map / Label-efficient and robust learning / Recognition and long-tail
Metric Embedded Discriminative Vocabulary Learning for High-Level Person Representation
Yang Yang, Zhen Lei, Shifeng Zhang, Hailin Shi, Stan Li
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 |
|---|---|---|---|
| learn discriminative vocabulary with metric-embedded term instead of k-means (linear coding) | KISSME on coded features · VIPeR | Rank-1 matching rate: 39.5% -> 41.1% | Improved |
| learn discriminative vocabulary instead of k-means | KISSME on coded features · PRID 450S | Rank-1 matching rate: 44.0% -> 45.9% | Improved |
| encode features with MED VL high-level coding vs original features | KISSME · VIPeR | Rank-1 matching rate: 38.9% -> 41.1% | Improved |
| encode features with MED VL high-level coding vs original features | KISSME · PRID 450S | Rank-1 matching rate: 42.6% -> 45.9% | Improved |
| use KISSME metric M in vocabulary objective instead of Euclidean M (ablation) | MED VL · VIPeR | Rank-1 matching rate: 39.5% -> 41.1% | Improved |
| use KISSME metric M in vocabulary objective instead of Euclidean M (ablation) | MED VL · PRID 450S | Rank-1 matching rate: 45.1% -> 45.9% | Improved |
| restrict coding to 5 nearest basis vectors (locality constraint) vs all 120 | SC/LLC with k-means vocabulary · VIPeR | Rank-1 matching rate: 39.5% -> 11.6% (SC) | Hurt |
| restrict coding to 5 nearest basis vectors (locality constraint) vs all 120 | SC/LLC with k-means vocabulary · PRID 450S | Rank-1 matching rate: 44.0% -> 8.5% (SC) | Hurt |
| vary vocabulary size from 80 to 300 (ablation) | MED VL · VIPeR | Rank-1 matching rate: 41.0% - 41.1% - 40.6% | No clear effect |
| vary vocabulary size from 80 to 300 (ablation) | MED VL · PRID 450S | Rank-1 matching rate: 45.7% - 45.9% - 45.8% | No clear effect |
Closest papers by meaning
- Embedding Deep Metric for Person Re-identification: A Study Against Large Variations (ECCV 2016 · similarity 0.86)
- Efficient Online Local Metric Adaptation via Negative Samples for Person Re-Identification (ICCV 2017 · similarity 0.85)
- Improving Deep Visual Representation for Person Re-identification by Global and Local Image-language Association (ECCV 2018 · similarity 0.85)
- Unsupervised Learning of Multi-Level Descriptors for Person Re-Identification (AAAI 2017 · similarity 0.85)
- DC-Former: Diverse and Compact Transformer for Person Re-identification (AAAI 2023 · similarity 0.84)
- Noisy-Correspondence Learning for Text-to-Image Person Re-identification (CVPR 2024 · similarity 0.83)
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.
