AI Paper Atlas

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

Towards Viewpoint-Invariant Visual Recognition via Adversarial Training

Shouwei Ruan, Yinpeng Dong, Hang Su, Jianteng Peng, Ning Chen, Xingxing Wei

ICCV 2023 · Adversarial robustness and security › Attacks on text and graphs · approach: Biologically inspired robust vision · 25 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
apply VIAT (adversarial viewpoint training) to a standard-trained classifierResNet-50 · IM3D, GMVFool viewpointstop-1 accuracy (%): 8.98 -> 59.61Improved
apply VIAT to a standard-trained classifierViT-B/16 · IM3D, GMVFool viewpointstop-1 accuracy (%): 29.10 -> 83.13Improved
replace ViewFool with GMVFool (Gaussian mixture) as inner maximization (ablation)VIAT ResNet-50 · IM3D, ViewFool viewpointstop-1 accuracy (%): 55.12 -> 59.84Improved
replace data augmentation with random viewpoints by VIAT (ablation)ResNet-50 · IM3D, GMVFool viewpointstop-1 accuracy (%): 33.52 -> 59.61Improved
apply VIAT on clean ImageNet subset accuracyViT-B/16 · ImageNet subset valtop-1 accuracy (%): 92.88 -> 92.56No clear effect
increase number of Gaussian components K in GMVFool (ablation)VIAT · IM3Drobust accuracy: K=1 entropy -10.17 vs K=5 -0.69; attack rate 92.13 vs 91.56Mixed
add distribution sharing across objects in same class (pi=0.5) (ablation)VIAT · IM3Dviewpoint robustnessImproved
evaluate VIAT-trained ResNet-50 on out-of-distribution datasetsResNet-50 · ImageNet-Vaccuracy (%): 28.15 -> 38.96Improved
evaluate VIAT-trained ResNet-50 on ImageNet-AResNet-50 · ImageNet-Aaccuracy (%): 19.03 -> 20.32Improved
use data augmentation with random viewpoint renderings on real OOD datasets (ablation)ResNet-50 · ImageNet-Aaccuracy (%): 19.03 -> 18.65No clear effect

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