Research map / Label-efficient and robust learning / Noisy labels
Noise-Resistant Deep Metric Learning With Ranking-Based Instance Selection
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 PRISM to MCL | MCL · CARS, symmetric noise 50% | Precision@1: 46.88 -> 72.93 | Helped |
| Add PRISM to MCL | MCL · CUB, symmetric noise 50% | Precision@1: 31.18 -> 56.03 | Helped |
| Add PRISM to MCL | MCL · CARS, Small Cluster noise 50% | Precision@1: 36.43 -> 68.26 | Helped |
| Add PRISM to MCL | MCL · CARS-98N | Precision@1: 38.73 -> 57.95 | Helped |
| Add PRISM to MCL | MCL · Food-101N | Precision@1: 52.58 -> 52.47 | No clear effect |
| Add PRISM to Soft Triple | Soft Triple · CARS-98N | Precision@1: 63.36 -> 64.81 | Helped |
| Use same-class minibatch similarities rather than memory-bank evidence for filtering (ablation) | PRISM with Soft Triple · CARS, Small Cluster noise 25% | Precision@1: 75.40 -> 74.66 | Hurt |
| Use same-class memory similarities without negative-class similarities (ablation) | PRISM with Soft Triple · CARS, Small Cluster noise 25% | Precision@1: 75.40 -> 75.12 | Hurt |
| Add memory-bank evidence to same-class minibatch filtering (ablation) | Noise filtering with Soft Triple · CARS, Small Cluster noise 25% | Precision@1: 74.66 -> 75.12 | Helped |
| Smooth the top-R threshold over a sliding window (ablation) | PRISM · CARS, Small Cluster noise 25% | Precision@1 | Helped |
| Set filtering rate R to 0.5 instead of R=0.0 (ablation) | MCL + PRISM · CARS-98N | Precision@1: 38.73 -> 57.95 | Helped |
| Use class centers rather than individual memory examples for filtering (ablation) | MCL + PRISM · SOP, symmetric noise 10%, 5,000 iterations | training time (Seconds): 12,294.76 -> 1,777.38 | Helped |
Closest papers by meaning
- Denoising Multi-Similarity Formulation: A Self-Paced Curriculum-Driven Approach for Robust Metric Learning (AAAI 2023 · similarity 0.88)
- Stochastic Class-Based Hard Example Mining for Deep Metric Learning (CVPR 2019 · similarity 0.87)
- Revisiting Training Strategies and Generalization Performance in Deep Metric Learning (ICML 2020 · similarity 0.86)
- Threshold-Consistent Margin Loss for Open-World Deep Metric Learning (ICLR 2024 · similarity 0.85)
- Handling Label Noise via Instance-Level Difficulty Modeling and Dynamic Optimization (NeurIPS 2025 · similarity 0.84)
- NoiseGPT: Label Noise Detection and Rectification through Probability Curvature (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.