Research map / Label-efficient and robust learning / Recognition and long-tail
Global and Local Mixture Consistency Cumulative Learning for Long-Tailed Visual Recognitions
Read the paperCheck 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 global and local mixture consistency (ablation) | GLMC training on CIFAR-100-LT (IF=100) · CIFAR-100-LT (IF=100) | Top-1 accuracy (%): 38.3 -> 50.11 | Helped |
| Add cumulative class-balanced reweighting (ablation) | GLMC training on CIFAR-100-LT (IF=100) · CIFAR-100-LT (IF=100) | Top-1 accuracy (%): 38.3 -> 44.63 | Helped |
| Add cumulative class-balanced reweighting to mixture consistency (ablation) | GLMC training on CIFAR-100-LT (IF=100) · CIFAR-100-LT (IF=100) | Top-1 accuracy (%): 50.11 -> 55.88 | Helped |
| Set mixture consistency weight γ from 0 to 10 (ablation) | GLMC on CIFAR-100-LT (IF=100) · CIFAR-100-LT (IF=100) | Top-1 accuracy (%): +2.41% | Helped |
| Increase rebalancing intensity through larger resampling weights (ablation) | GLMC first-stage training · CIFAR-100-LT (IF=100) | Top-1 accuracy (%) | Hurt |
| Use label reweighting for long-tailed recognition (ablation) | GLMC first-stage training · CIFAR-100-LT (IF=100) | Top-1 accuracy (%) | Helped |
| Finetune the classifier with MaxNorm | GLMC · CIFAR-100-LT (IF=100, 50, 10) | Top-1 accuracy (%): 55.88 -> 57.11; 61.08 -> 62.32; 70.74 -> 72.33 | Mixed |
| Finetune the classifier with MaxNorm | GLMC · ImageNet-LT (Many) | Top-1 accuracy (%): 70.1 -> 60.8 | Hurt |
| Finetune the classifier with MaxNorm | GLMC · ImageNet-LT (Few) | Top-1 accuracy (%): 30.4 -> 45.5 | Helped |
| Add global and local mixture consistency to supervised training | GLMC on full ImageNet · ImageNet (full) | Top-1 accuracy (%): 76.4 -> 80.2 | Helped |
| Add global and local mixture consistency to supervised training | GLMC on full CIFAR · CIFAR-10 (full), CIFAR-100 (full) | Top-1 accuracy (%): 94.85 -> 97.23; 75.28 -> 83.05 | Helped |
Closest papers by meaning
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.89)
- Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition (NeurIPS 2024 · similarity 0.88)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- Improving Visual Prompt Tuning by Gaussian Neighborhood Minimization for Long-Tailed Visual Recognition (NeurIPS 2024 · similarity 0.88)
- Decoupled Contrastive Learning for Long-Tailed Recognition (AAAI 2024 · similarity 0.87)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 2024 · similarity 0.87)
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.