Research map / Label-efficient and robust learning / Domain adaptation and generalization
AdaTransform: Adaptive Data Transformation
Zhiqiang Tang, Xi Peng, Tingfeng Li, Yizhe Zhu, Dimitris N. Metaxas
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 AdaCutout instead of fixed Cutout | ResNet32 · CIFAR-10 (10% train) | accuracy: 77.21 -> 78.02 | Improved |
| add AdaCutout instead of fixed Cutout | ResNet32 · CIFAR-100 (10% train) | accuracy: 40.41 -> 41.02 | Improved |
| add AdaErasing instead of random Erasing | ResNet32 · CIFAR-10 (10% train) | accuracy: 77.25 -> 78.12 | Improved |
| add AdaErasing instead of random Erasing | ResNet32 · CIFAR-100 (10% train) | accuracy: 40.53 -> 41.21 | Improved |
| use adaptive texture transformation instead of random augmentation (ablation) | two stacked hourglasses · MPII Human Pose | mean PCKh: 88.1 -> 88.7 | Improved |
| use adaptive cutout instead of random augmentation (ablation) | two stacked hourglasses · MPII Human Pose | mean PCKh: 88.1 -> 88.6 | Improved |
| use adaptive scale and rotation instead of random augmentation (ablation) | two stacked hourglasses · MPII Human Pose | mean PCKh: 88.1 -> 88.9 | Improved |
| combine all adaptive transformations (AdaAll) (ablation) | two stacked hourglasses · MPII Human Pose | mean PCKh: 88.1 -> 89.3 | Improved |
| use AdaImgTexture instead of random augmentation under texture perturbation | two stacked hourglasses · MPII Human Pose (texture-perturbed test) | mean PCKh: 86.3 -> 87.8 | Improved |
| use AdaTransform instead of AdvAug | two stacked hourglasses · 300-W Full Set | NME: 3.28 -> 3.24 | Improved |
| use AdaTransform instead of AdvAug | two stacked hourglasses · LSP test | mean PCK: 92.5 -> 92.9 | Improved |
| use AdaTransform instead of AutoAugment | Shake-Shake · CIFAR-10 (full train) | error (%): 1.99 -> 2.11 | Hurt |
Closest papers by meaning
- Tradeoffs in Data Augmentation: An Empirical Study (ICLR 2021 · similarity 0.86)
- Regularizing Deep Networks with Semantic Data Augmentation (TPAMI 2021 · similarity 0.85)
- A Group-Theoretic Framework for Data Augmentation (NeurIPS 2020 · similarity 0.85)
- AdaXpert: Adapting Neural Architecture for Growing Data (ICML 2021 · similarity 0.84)
- AdvST: Revisiting Data Augmentations for Single Domain Generalization (AAAI 2024 · similarity 0.83)
- First-Order Manifold Data Augmentation for Regression Learning (ICML 2024 · similarity 0.83)
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.
