AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Can Shape Structure Features Improve Model Robustness Under Diverse Adversarial Settings?

Mingjie Sun, Zichao Li, Chaowei Xiao, Haonan Qiu, Bhavya Kailkhura, Mingyan Liu, Bo Li

ICCV 2021 · Adversarial robustness and security › Certified and empirical robustness · approach: Robustness-accuracy trade-off analysis · 19 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
replace Canny with Robust Canny (add noise masking by gradient-magnitude thresholding) (ablation)EdgeNetRob · Fashion MNIST, PGD-40robust accuracy: 39.99 -> 76.75Improved
replace RCF edge detector with Canny (ablation)EdgeNetRob · Fashion MNIST, PGD-40robust accuracy: 0.18 -> 39.99Improved
replace Canny with Robust Canny (ablation)EdgeNetRob · Fashion MNISTclean accuracy: 88.32 -> 87.00Hurt
add GAN texture refilling (EdgeGANRob vs EdgeNetRob)EdgeNetRob · Fashion MNISTclean accuracy: 87.00 -> 87.14No clear effect
add GAN texture refilling (EdgeGANRob vs EdgeNetRob)EdgeNetRob · CelebAclean accuracy: 94.51 -> 95.88Improved
add GAN texture refilling (EdgeGANRob vs EdgeNetRob)EdgeNetRob · CIFAR-10clean accuracy: 79.21 -> 76.25Hurt
add GAN texture refilling (EdgeGANRob vs EdgeNetRob)EdgeNetRob · Tiny ImageNet, PGD-40robust accuracy: 19.53 -> 13.55Hurt
add CNN-based edges to Robust Canny (Combined Edge) (ablation)EdgeNetRob · CIFAR-10clean and PGD-40 accuracy: 67.85 -> 79.21 clean; 36.31 -> 33.08 robustMixed
use EdgeNetRob instead of PGD adversarial trainingEdgeNetRob · Fashion MNIST, PGD-40robust accuracy: 72.62 -> 76.75Improved
use EdgeNetRob instead of PGD adversarial trainingEdgeNetRob · CIFAR-10, PGD-40robust accuracy: 44.15 -> 33.08Hurt
use EdgeNetRob instead of a vanilla networkEdgeNetRob · CIFAR-10 backdoor (Pixel, 5% poison)poison attack success rate: 99.50 -> 9.42Improved
use EdgeGANRob instead of Spectral SignatureEdgeGANRob · Fashion MNIST backdoor (Pixel, 20% poison)poison attack success rate: 45.62 -> 0.07Improved

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