AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Improving Robustness Against Common Corruptions With Frequency Biased Models

Tonmoy Saikia, Cordelia Schmid, Thomas Brox

ICCV 2021 · Adversarial robustness and security › Attacks on text and graphs · approach: Biologically inspired robust vision · 51 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 TV regularization on first conv feature maps to AugMix (ablation)AugMix ResNet18 · ImageNet-C-100mCE: 40.9 -> 35.9Improved
add TV regularization on first conv feature maps to AugMix (ablation)AugMix ResNet18 · ImageNet-100 cleanclean error: 11.8 -> 14.8Hurt
finetune with Gaussian noise and blur (HF bias) (ablation)AugMix ResNet18 · ImageNet-C-100mCE: 40.9 -> 32.5Improved
finetune with Gaussian noise and blur (HF bias) (ablation)AugMix ResNet18 · ImageNet-100 cleanclean error: 11.8 -> 13.2Hurt
finetune with contrast augmentation (LF bias) (ablation)AugMix ResNet18 · ImageNet-C-100mCE: 40.9 -> 39.1Improved
add TV regularization to HF-finetuned model (ablation)AM-HF · ImageNet-C-100mCE: 32.5 -> 31.5Improved
use HF and LF experts (different biases) instead of same-bias ensemble (ablation)AM ensemble · ImageNet-C-100mCE: 39.1 -> 25.9Improved
use HF+LF experts instead of ensemble of two same-bias augmented models (ablation)AM(Gauss,Cont) ensemble · ImageNet-C-100mCE: 29.0 -> 25.9Improved
replace AMDA ensemble with RoHL experts (ablation)Ensemble(AMDA, AMDA) · ImageNet-CmCE: 51.9 -> 47.9Improved
add RoHL HF/LF experts to AMDAAMDA · ImageNet cleanclean error: 24.2 -> 22.7Improved
apply TV regularization to deeper layers instead of conv1 (ablation)TV regularization ResNet18 · ImageNet-C-100high-frequency corruption errorHurt
use larger TV factor lambda (>=1e-4) (ablation)TV regularization · ImageNet-100clean and corruption error: 1e-6 to 9e-5 good; larger degradesHurt

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