Research map / Label-efficient and robust learning / Recognition and long-tail
Augmentation-invariant Learning Strategy via Data Augmentation for Improving Model Generalization
Yu Miao, Juanjuan Zhao, Sijie Song, Ran Gong, Yuanqian Zhu, Lusha Qi, Yan Qiang
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 AILS to MobileNetV3 | MobileNetV3 · MedMNIST+ 2D (backbone study) | accuracy (%): 70.1 -> 76.3 | Improved |
| add AILS to ResNeSt-101 | ResNeSt-101 · MedMNIST+ 2D (backbone study) | accuracy (%): 80.3 -> 85.1 | Improved |
| add AILS to Swin-Tiny | Swin-Tiny · MedMNIST+ 2D (backbone study) | accuracy (%): 80.5 -> 84.4 | Improved |
| add AILS reduces epochs to convergence | ResNeSt-101 · MedMNIST+ 2D (backbone study) | epochs to convergence: 65 -> 53 | Improved |
| replace BSDA with AILS | AILS vs BSDA · PneumoniaMNIST | accuracy (%): 88.8 -> 92.2 | Improved |
| replace BSDA with AILS | AILS vs BSDA · TissueMNIST | accuracy (%): 73.3 -> 79.3 | Improved |
| replace BSDA with AILS | AILS vs BSDA · BloodMNIST | accuracy (%): 98.8 -> 98.9 | No clear effect |
| use AILS with 20% training data instead of BSDA | AILS vs BSDA · MedMNIST+ 2D avg (20% training data) | accuracy (%): 67.2 -> 69.4 | Improved |
| use AILS with 50% training data instead of BSDA | AILS vs BSDA · MedMNIST+ 2D avg (50% training data) | accuracy (%): 74.1 -> 81.8 | Improved |
| lower augmentation intensity alpha from 0.3 to 0.1 (ablation) | AILS · MedMNIST+ 2D (9 subsets, avg) | accuracy (%): 85.74 -> 79.84 | Hurt |
| raise augmentation intensity alpha from 0.3 to 0.7 (ablation) | AILS · MedMNIST+ 2D (9 subsets, avg) | accuracy (%): 85.74 -> 76.04 | Hurt |
| lower KL weight beta from 1.0 to 0.1 (ablation) | AILS · MedMNIST+ 2D (9 subsets, avg) | accuracy (%): 85.74 -> 78.24 | Hurt |
Closest papers by meaning
- Rethinking Data Augmentation for Single-Source Domain Generalization in Medical Image Segmentation (AAAI 2023 · similarity 0.88)
- Adversarial Data Augmentation via Deformation Statistics (ECCV 2020 · similarity 0.86)
- A Textbook Remedy for Domain Shifts: Knowledge Priors for Medical Image Analysis (NeurIPS 2024 · similarity 0.86)
- Training Like a Medical Resident: Context-Prior Learning Toward Universal Medical Image Segmentation (CVPR 2024 · similarity 0.85)
- TinyMIG: Transferring Generalization from Vision Foundation Models to Single-Domain Medical Imaging (ICML 2025 · similarity 0.84)
- Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging (AAAI 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.
