Research map / Label-efficient and robust learning / Recognition and long-tail
SelectAugment: Hierarchical Deterministic Sample Selection for Data Augmentation
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 |
|---|---|---|---|
| Use deterministic hierarchical sample selection for AutoAugment instead of augmenting all samples | AutoAugment with WRN-28-10 · CIFAR-10 | test accuracy: 97.28 -> 97.65 | Helped |
| Use deterministic hierarchical sample selection for AutoAugment instead of augmenting all samples | AutoAugment with ResNet-50 · ImageNet validation | Top-1 accuracy: 77.61 -> 78.16 | Helped |
| Add a child policy to the parent policy, enabling instance-level selection rather than random sample allocation (ablation) | SelectAugment with Mixup · CIFAR-10 | test accuracy: 97.05 -> 97.33 | Helped |
| Add a parent policy to the child policy, choosing a batch-level augmentation ratio (ablation) | SelectAugment with Mixup · CIFAR-10 | test accuracy: 97.21 -> 97.33 | Helped |
| Add a child policy to the parent policy, enabling instance-level selection rather than random sample allocation (ablation) | SelectAugment with Mixup · CIFAR-100 | test accuracy: 82.97 -> 83.37 | Helped |
| Add a parent policy to the child policy, choosing a batch-level augmentation ratio (ablation) | SelectAugment with Mixup · ImageNet validation | Top-5 accuracy: 77.34 -> 77.84 | Helped |
| Use an augmentation-ratio pool with interval number 10 rather than interval number 100 (ablation) | SelectAugment with Mixup · CIFAR-100 | test accuracy: 83.15 -> 83.37 | Helped |
| Select an augmentation operation for each selected sample in addition to selecting samples | SelectAugment+ with WRN-28-10 · CIFAR-100 | test accuracy: 83.81 -> 84.11 | Helped |
Closest papers by meaning
- MetaAugment: Sample-Aware Data Augmentation Policy Learning (AAAI 2021 · similarity 0.90)
- Optimizing Data Augmentation through Bayesian Model Selection (ICLR 2026 · similarity 0.88)
- AutoAugment: Learning Augmentation Strategies From Data (CVPR 2019 · similarity 0.88)
- When Dynamic Data Selection Meets Data Augmentation: Achieving Enhanced Training Acceleration (ICML 2025 · similarity 0.87)
- Revisiting Data Augmentation in Deep Reinforcement Learning (ICLR 2024 · similarity 0.86)
- A CLIP-Powered Framework for Robust and Generalizable Data Selection (ICLR 2025 · 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.