Research map / Label-efficient and robust learning / Recognition and long-tail
Hypergraph-Induced Semantic Tuplet Loss for Deep Metric Learning
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 |
|---|---|---|---|
| Use HIST instead of IBC | ResNet-50 metric embedding · CUB-200-2011 | R@1: 70.3 -> 71.4±0.2 | Helped |
| Use HIST instead of IBC | ResNet-50 metric embedding · CARS-196 | R@1: 88.1 -> 89.6±0.2 | Helped |
| Use HIST instead of IBC | ResNet-50 metric embedding · SOP | R@1: 81.4 -> 81.4±0.2 | No clear effect |
| Add individual sample classification to LD-only (ablation) | CARS-196 embedding model · CARS-196 | R@1: 87.3 ± 0.4 -> 86.4 ± 0.2 | Hurt |
| Replace individual classification with Transformer-like pairwise classification (ablation) | CARS-196 embedding model · CARS-196 | R@1: 86.4 ± 0.2 -> 87.8 ± 0.3 | Helped |
| Replace individual classification with IBC graph classification (ablation) | CARS-196 embedding model · CARS-196 | R@1: 86.4 ± 0.2 -> 87.6 ± 0.3 | Helped |
| Use hypergraph classification with semantic tuplets rather than IBC graph classification (ablation) | CARS-196 embedding model · CARS-196 | R@1: 87.6 ± 0.3 -> 89.6 ± 0.2 | Helped |
| Add negative samples to positive-only hyperedges (ablation) | H-Pos on CARS-196 · CARS-196 | R@1: 87.4 ± 0.2 -> 89.6 ± 0.2 | Helped |
| Add distribution loss LD (ablation) | HIST on CARS-196 · CARS-196 | R@1: 88.3 ± 0.2 -> 89.6 ± 0.2 | Helped |
| Increase mini-batch size from 32 to 48 (ablation) | HIST on CARS-196 · CARS-196 | Recall@1: 87.3 -> 88.0 | Helped |
| Increase HGNN depth from 2 to 4 layers (ablation) | HIST on CARS-196 · CARS-196 | Recall@1 | Hurt |
| Set the negative-sample scaling factor α to 0 instead of a positive value (ablation) | HIST on CARS-196 · CARS-196 | Recall@1 | Hurt |
Closest papers by meaning
- Deep Metric Learning with Graph Consistency (AAAI 2021 · similarity 0.85)
- Towards Hierarchy–Uniformity Equilibrium: Recovering Semantic Depth in Hypergraph Contrastive Learning (ICML 2026 · similarity 0.85)
- Ranked List Loss for Deep Metric Learning (CVPR 2019 · similarity 0.85)
- Deep Metric Learning Beyond Binary Supervision (CVPR 2019 · similarity 0.85)
- Hypergraph Vision Transformers: Images are More than Nodes, More than Edges (CVPR 2025 · similarity 0.83)
- Learning Structured Representations with Hyperbolic Embeddings (NeurIPS 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.