Research map / Architectures and efficiency / Neural architecture search
AutoAugment: Learning Augmentation Strategies From Data
Check 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 |
|---|---|---|---|
| Add the learned AutoAugment policy to baseline augmentation and Cutout | PyramidNet+ShakeDrop · CIFAR-10 | error rate (%): 2.3 -> 1.5 ± 0.1 | Helped |
| Add the learned policy to baseline augmentation and Cutout | Shake-Shake (26 2x96d) · Reduced CIFAR-10 | error rate (%): 13.4 -> 10.0 ± 0.2 | Helped |
| Add the learned policy to baseline augmentation and Cutout | PyramidNet+ShakeDrop · CIFAR-100 | error rate (%): 12.2 -> 10.7 ± 0.2 | Helped |
| Add the learned policy to baseline augmentation and Cutout | Shake-Shake (26 2x96d) · SVHN | error rate (%): 1.2 -> 1.0 | Helped |
| Add the learned policy without Cutout | Shake-Shake (26 2x96d) · Reduced SVHN | error rate (%): 12.3 -> 5.9 | Helped |
| Add Cutout to baseline augmentation (ablation) | Wide-ResNet-28-10 · Reduced SVHN | error rate (%): 13.2 -> 32.5 | Hurt |
| Add AutoAugment to Inception-style preprocessing | AmoebaNet-C (6,228) · ImageNet validation | Top-1 accuracy (%): 83.1 -> 83.5 | Helped |
| Transfer the ImageNet-learned policy | Inception v4 · Oxford 102 Flowers | Top-1 error rate (%): 6.7 -> 4.6 | Helped |
| Transfer the ImageNet-learned policy | Inception v4 · Caltech-101 | Top-1 error rate (%): 19.4 -> 13.1 | Helped |
| Transfer the ImageNet-learned policy | Inception v4 · Oxford-IIIT Pets | Top-1 error rate (%): 13.5 -> 11.0 | Helped |
| Transfer the ImageNet-learned policy | Inception v4 · FGVC Aircraft | Top-1 error rate (%): 9.1 -> 7.3 | Helped |
| Transfer the ImageNet-learned policy | Inception v4 · Stanford Cars | Top-1 error rate (%): 6.4 -> 5.2 | Helped |
Closest papers by meaning
- Adversarial AutoAugment (ICLR 2020 · similarity 0.92)
- RandAugment: Practical Automated Data Augmentation with a Reduced Search Space (NeurIPS 2020 · similarity 0.92)
- MetaAugment: Sample-Aware Data Augmentation Policy Learning (AAAI 2021 · similarity 0.90)
- Fast AutoAugment (NeurIPS 2019 · similarity 0.89)
- AdaAug: Learning Class- and Instance-adaptive Data Augmentation Policies (ICLR 2022 · similarity 0.88)
- AutoDO: Robust AutoAugment for Biased Data With Label Noise via Scalable Probabilistic Implicit Differentiation (CVPR 2021 · similarity 0.88)
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.