AI Paper Atlas

Research map / Detection and segmentation / CNN features

Content-Aware Token Sharing for Efficient Semantic Segmentation With Vision Transformers

Chenyang Lu, Daan de Geus, Gijs Dubbelman

CVPR 2023 · Detection and segmentation › CNN features · approach: Plain and high-resolution ViTs · 35 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
Share tokens for predicted semantically uniform neighboring patches instead of processing every patch separatelySegmenter with ViT-S/16 · ADE20K valmIoU: 45.0 -> 45.1No clear effect
Share tokens for predicted semantically uniform neighboring patches instead of processing every patch separatelySegmenter with ViT-S/16 · ADE20K valIm/sec: 122 -> 162Improved
Select shared patches with CTS rather than randomly (ablation)Segmenter with ViT-S/16 · ADE20K valmIoU: 42.9 -> 45.1Improved
Use a larger PolicyNet (ablation)CTS with Segmenter, ViT-S/16 · ADE20K valmIoU: 45.1 -> 45.3Improved
Use a smaller custom PolicyNet (ablation)CTS with Segmenter, ViT-S/16 · ADE20K valmIoU: 45.1 -> 44.0Hurt
Generate sharing policy from segmentation predictions rather than the class-agnostic PolicyNet (ablation)CTS with Segmenter, ViT-S/16 · ADE20K valmIoU: 45.1 -> 44.3; 45.1 -> 44.8Hurt
Use a ground-truth segmentation policy instead of predicted PolicyNet policy (ablation)CTS with Segmenter, ViT-S/16 · ADE20K valmIoU: 45.1 -> 47.2Improved
Choose shared patches randomly instead of using a content-aware policy (ablation)CTS with Segmenter, ViT-S/16 · ADE20K valmIoU: 45.0 -> 42.9Hurt
Apply CTS with 30% token reductionSegmenter with ViT-B/16 · ADE20K valmIoU: 48.5 -> 48.7No clear effect
Apply CTS with 30% token reductionBEiT-L + UPerNet · ADE20K valIm/sec: 3.3 -> 5.4Improved
Apply CTS with 30% token reductionBEiT-L + UPerNet · ADE20K valmIoU: 56.8 -> 56.6No clear effect
Apply CTS with 30% token reductionSegmenter with ViT-S/16 · Pascal Context validationmIoU: 53.0 -> 52.9No 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