Research map / Label-efficient and robust learning / Recognition and long-tail
Imagine by Reasoning: A Reasoning-Based Implicit Semantic Data Augmentation for Long-Tailed Classification
Xiaohua Chen, Yucan Zhou, Dayan Wu, Wanqian Zhang, Yu Zhou, Bo Li, Weiping Wang
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 reasoning-based augmentation to transfer semantic transformations from similar classes (ablation) | RISDA · CIFAR-100-LT λ=200 | error rate (%): 56.97 -> 55.24 | Improved |
| Add reasoning-based augmentation to transfer semantic transformations from similar classes (ablation) | RISDA · CIFAR-100-LT λ=100 | error rate (%): 51.09 -> 49.84 | Improved |
| Add reasoning-based augmentation to transfer semantic transformations from similar classes (ablation) | RISDA · CIFAR-100-LT λ=50 | error rate (%): 47.60 -> 46.16 | Improved |
| Add reasoning-based augmentation to transfer semantic transformations from similar classes (ablation) | RISDA · CIFAR-100-LT λ=20 | error rate (%): 42.60 -> 41.33 | Improved |
| Add reasoning-based augmentation to transfer semantic transformations from similar classes (ablation) | RISDA · CIFAR-100-LT λ=10 | error rate (%): 38.43 -> 37.62 | Improved |
| Add class re-weighting (ablation) | RISDA · CIFAR-100-LT λ=200 | error rate (%): 36.82 -> 55.24 | Hurt |
| Add class re-weighting (ablation) | RISDA · CIFAR-100-LT λ=100 | error rate (%): 42.76 -> 49.84 | Hurt |
| Add class re-weighting (ablation) | RISDA · CIFAR-100-LT λ=50 | error rate (%): 46.98 -> 46.16 | Hurt |
| Add class re-weighting (ablation) | RISDA · CIFAR-100-LT λ=20 | error rate (%): 54.40 -> 41.33 | Improved |
| Add class re-weighting (ablation) | RISDA · CIFAR-100-LT λ=10 | error rate (%): 58.60 -> 37.62 | Improved |
| Transfer covariance from similar classes rather than using only class-specific covariance (ablation) | Implicit augmentation on CIFAR-100-LT · CIFAR-100-LT λ=200, 100, 50, 20, 10 | error rate (%): 56.97, 51.09, 47.6, 42.6, 38.43 -> 55.24, 49.84, 46.16, 41.33, 37.62 | Improved |
| Use RISDA instead of MetaSAug | Long-tailed classification · ImageNet-LT | error rate (%): 52.61 -> 49.31 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MetaSAug — MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition (CVPR 2021 · similarity 0.89)
- Long-Tail Recognition via Compositional Knowledge Transfer (CVPR 2022 · similarity 0.88)
- Balancing Logit Variation for Long-Tailed Semantic Segmentation (CVPR 2023 · similarity 0.87)
- Learning from Neighbors: Category Extrapolation for Long-Tail Learning (CVPR 2025 · similarity 0.86)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.85)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 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.
