AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Feature Purification Matters: Suppressing Outlier Propagation for Training-Free Open-Vocabulary Semantic Segmentation

Shuo Jin, Siyue Yu, Bingfeng Zhang, Mingjie Sun, Yi Dong, Jimin Xiao

ICCV 2025 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary semantic segmentation · 7 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
add Self-adaptive Outlier Mitigator (SOM) (ablation)ClearCLIP baseline · VOC21mIoU: 57.0 -> 60.1Improved
add SOM (ablation)ClearCLIP baseline · ADE20KmIoU: 16.7 -> 18.2Improved
add Semantic-aware Attention Enhancer (SAE) after SOM (ablation)SFP · VOC21mIoU: 60.1 -> 62.3Improved
add SAE after SOM (ablation)SFP · ADE20KmIoU: 18.2 -> 19.5Improved
add Hierarchical Attention Integrator (HAI) (ablation)SFP · VOC21mIoU: 62.3 -> 63.9Improved
add HAI (ablation)SFP · ADE20KmIoU: 19.5 -> 20.8Improved
integrate attention maps from all layers (l0=1) instead of starting at layer 11 (ablation)SFP HAI · VOC21mIoU: 62.5 -> 63.9Improved
integrate attention maps from all layers (l0=1) instead of starting at layer 11 (ablation)SFP HAI · ADE20KmIoU: 19.7 -> 20.8Improved
start HAI at layer 5 instead of layer 1 (ablation)SFP HAI · VOC21 and ADE averagemIoU: 42.4 -> 42.0Hurt
swap SFP for ProxyCLIP with ViT-L/14 backboneSFP vs ProxyCLIP · ADE20KmIoU: 22.1 vs 22.6Hurt
use SFP instead of NACLIPNACLIP · Average of 8 datasetsmIoU: 39.4 -> 44.0Improved
use SFP instead of ProxyCLIP (extra DINO backbone)ProxyCLIP · CityscapesmIoU: 38.1 -> 41.1Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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