Research map / Detection and segmentation / CNN features
Sample-Aware RandAugment: Search-Free Automatic Data Augmentation for Effective Image Recognition
Anqi Xiao, Weichen Yu, Hongyuan Yu
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 scaling factor gamma to MIS (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.49 -> 84.64 | Improved |
| add random augmentation in Step 1 exploration (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.45 -> 84.64 | Improved |
| add distribution exploration (Step 1) alongside refinement (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.09 -> 84.64 | Improved |
| add sample-aware refinement (Step 2 & 3) alongside exploration (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 83.93 -> 84.64 | Improved |
| split exploration and refinement into separate sub-batches (asymmetric) (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.54 -> 84.64 | Improved |
| use cosine similarity instead of Euclidean distance for MIS (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.37 -> 84.64 | Improved |
| use cosine similarity instead of Jaccard similarity for MIS (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.03 -> 84.64 | Improved |
| augmentation depth D=2 instead of D=4 (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.02 -> 84.64 | Improved |
| normalization scale epsilon 2 instead of 0 (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.44 -> 84.64 | Improved |
| add TranslateY operator to the candidate set (ablation) | SRA · CIFAR-100 (WRN-28-10) | top-1 accuracy (%): 84.11 -> 84.64 | Improved |
| use SRA instead of RA (2, 9) | ResNet-50 · ImageNet-LT (overall) | top-1 accuracy (%): 45.18 -> 46.76 | Improved |
| use SRA instead of RA (2, 9) | ResNet-50 · Food101 | top-1 accuracy (%): 85.97 -> 87.04 | Improved |
Closest papers by meaning
- RandAugment: Practical Automated Data Augmentation with a Reduced Search Space (NeurIPS 2020 · similarity 0.93)
- MetaAugment: Sample-Aware Data Augmentation Policy Learning (AAAI 2021 · similarity 0.90)
- Faster AutoAugment: Learning Augmentation Strategies Using Backpropagation (ECCV 2020 · similarity 0.89)
- AutoAugment: Learning Augmentation Strategies From Data (CVPR 2019 · similarity 0.89)
- Scale-Aware Automatic Augmentations for Object Detection With Dynamic Training (TPAMI 2022 · similarity 0.89)
- When Dynamic Data Selection Meets Data Augmentation: Achieving Enhanced Training Acceleration (ICML 2025 · similarity 0.86)
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.
