AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Adversarial Normalization: I Can Visualize Everything (ICE)

Hoyoung Choi, Seungwan Jin, Kyungsik Han

CVPR 2023 · Architectures and efficiency › Attention and transformers · approach: Token merging and masked pretraining · 6 citations (OpenAlex)

Read the paper

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 patch-wise classification with adversarial normalizationDeiT-S · ImageNet-SegmentationPixel-wise accuracy: 4.05%Improved
Add patch-wise classification with adversarial normalizationDeiT-S · ImageNet-SegmentationMean IoU: 3.94%Improved
Add patch-wise classification with adversarial normalizationDeiT-S · PascalVOC12Jaccard similarity: 41.32Improved
Use ICE instead of DINO attention for foreground/background separationDeiT-S · PascalVOC12Jaccard similarity: 40.30 -> 41.32Improved
Add patch-wise classification with adversarial normalizationDeiT-S · CUB-200-2011GT Loc: 92.5Improved
Add patch-wise classification with adversarial normalizationDeiT-S · PascalVOC07 and PascalVOC12CorLoc: 62.2; 66.0Mixed
Use ICE for patch selection instead of EViTDeiT-S, keep-rate 0.5 · ImageNetTop-1 accuracy (%): 78.39 -> 78.49Improved
Use ICE for patch selection instead of EViTDeiT-S, keep-rate 0.5 · ImageNetThroughput: 1701 -> 1547Hurt
Freeze DeiT-S parameters and train only the two MLP layers (ablation)ICE · ImageNet-SegmentationMean IoU: 67.94 -> 62.12Hurt
Add background scale α=0.5 rather than use α=1.0 (ablation)ICE with DeiT-S · ImageNet-SegmentationPixel-wise accuracy: 63.45 -> 84.85Improved
Add background scale α=0.5 rather than use α=1.0 (ablation)ICE with DeiT-S · ImageNet-SegmentationMean IoU: 45.36 -> 67.94Improved
Use class weight λclass=0.975 rather than λclass=0.5 (ablation)ICE with DeiT-S · ImageNet-SegmentationMean IoU: 57.59 -> 67.94Improved

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