AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Circumventing Outliers of AutoAugment with Knowledge Distillation

Longhui Wei, An Xiao, Lingxi Xie, Xiaopeng Zhang, Xin Chen, Qi Tian

ECCV 2020 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 69 citations (OpenAlex)

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 changedWhereMeasuredEffect
add knowledge distillation (M=1) (ablation)RandAugment · CIFAR-100 modified augment spaceerror rate: 19.5 -> 17.6Improved
add knowledge distillation (M=7) (ablation)RandAugment · CIFAR-100 modified augment spaceerror rate: 55.1 -> 48.0Improved
add knowledge distillation (M=0, no augmentation) (ablation)RandAugment · CIFAR-100 modified augment spaceerror rate: 18.4 -> 18.0Improved
increase distortion magnitude M from 0 to 7 without KD (ablation)RandAugment · CIFAR-100 modified augment spaceerror rate: 18.4 -> 55.1Hurt
add KD with Wide-ResNet-28-10 teacher to AutoAugmentAutoAugment · CIFAR-100 Wide-ResNet-28-10error rate: 17.1 -> 16.2Improved
add KD with Wide-ResNet-28-10 teacher to AutoAugmentAutoAugment · CIFAR-100 Shake-Shake (26 2x96D)error rate: 14.3 -> 13.8Improved
add KD with itself as teacher to AutoAugmentAutoAugment · CIFAR-100 PyramidNet+ShakeDroperror rate: 10.7 -> 10.6Improved
add KD with weaker teacher to AutoAugmentAutoAugment · CIFAR-10 Shake-Shake (26 2x96D)error rate: 2.0 -> 1.8Improved
add KD to AutoAugmentAutoAugment · CIFAR-10 PyramidNet+ShakeDroperror rate: 1.5 -> 1.5No clear effect
add KD during the extra 50 clean-image epochs (ablation)AutoAugment Wide-ResNet-28-10 · CIFAR-100error rate: 16.8 -> 16.2Improved
use KD (lambda 0.1) on EfficientNet-B0 with M=9 (ablation)RandAugment · ImageNettop-1 accuracy: +0.3%Improved
increase K from 5 to larger values (ablation)KD · ImageNet EfficientNet-B0top-1 accuracyHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy