AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

An Impartial Take to the CNN vs Transformer Robustness Contest

Francesco Pinto, Philip H. S. Torr, Puneet K. Dokania

ECCV 2022 · Architectures and efficiency › Attention and transformers · approach: Efficient and robust vision transformers · 55 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
swap CNN for transformer architecture (BiT-R152x4 to ViT-B)ViT-B/16 · ImageNet-O OoDAUROC: 80.07 -> 79.89Hurt
swap Swin-L for ConvNeXt-LConvNeXt-L · ImageNet-O OoDAUROC: 87.76 -> 89.07Improved
swap top CNN for top transformerViT-L/16 · ImageNet-O OoDAUROC: 90.04 -> 90.60No clear effect
flip positive class and use AUPR instead of AUROC (ablation)ConvNeXt-XL · ImageNet-O OoDAUPR: 97.19 -> 68.50Mixed
rebalance in/out samples (ablation)ViT-B/16 · ImageNet-O OoDAUPR: 36.77 -> 73.77Mixed
use ViT-B/16 instead of Swin-BViT-B/16 · ImageNet-1KECE: 8.40 -> 1.40Improved
swap ConvNeXt-XL for ViT-L16 under ImageNet-R shiftViT-L16 · ImageNet-RECE: 2.92 -> 3.07Hurt
scale up BiT-R50x1 to BiT-R152x4BiT · ImageNet-1KECE: 3.49 -> 9.38Hurt
use ViT-L/16 instead of ViT-B/16 for simplicity biasViT-L/16 · R-CIFARaccuracy: 99.79 -> 95.66Hurt
swap ViT-L16 for Swin-L under ImageNet-RSwin-L · ImageNet-RPRR: -34.12 -> 36.53Improved

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