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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Self-adaptive Outlier Mitigator (SOM) (ablation) | ClearCLIP baseline · VOC21 | mIoU: 57.0 -> 60.1 | Improved |
| add SOM (ablation) | ClearCLIP baseline · ADE20K | mIoU: 16.7 -> 18.2 | Improved |
| add Semantic-aware Attention Enhancer (SAE) after SOM (ablation) | SFP · VOC21 | mIoU: 60.1 -> 62.3 | Improved |
| add SAE after SOM (ablation) | SFP · ADE20K | mIoU: 18.2 -> 19.5 | Improved |
| add Hierarchical Attention Integrator (HAI) (ablation) | SFP · VOC21 | mIoU: 62.3 -> 63.9 | Improved |
| add HAI (ablation) | SFP · ADE20K | mIoU: 19.5 -> 20.8 | Improved |
| integrate attention maps from all layers (l0=1) instead of starting at layer 11 (ablation) | SFP HAI · VOC21 | mIoU: 62.5 -> 63.9 | Improved |
| integrate attention maps from all layers (l0=1) instead of starting at layer 11 (ablation) | SFP HAI · ADE20K | mIoU: 19.7 -> 20.8 | Improved |
| start HAI at layer 5 instead of layer 1 (ablation) | SFP HAI · VOC21 and ADE average | mIoU: 42.4 -> 42.0 | Hurt |
| swap SFP for ProxyCLIP with ViT-L/14 backbone | SFP vs ProxyCLIP · ADE20K | mIoU: 22.1 vs 22.6 | Hurt |
| use SFP instead of NACLIP | NACLIP · Average of 8 datasets | mIoU: 39.4 -> 44.0 | Improved |
| use SFP instead of ProxyCLIP (extra DINO backbone) | ProxyCLIP · Cityscapes | mIoU: 38.1 -> 41.1 | Improved |
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.
- ProxyCLIP — ProxyCLIP: Proxy Attention Improves CLIP for Open-Vocabulary Segmentation (ECCV 2024) · better on 2 of 2 shared results
- ClearCLIP — ClearCLIP: Decomposing CLIP Representations for Dense Vision-Language Inference (ECCV 2024) · better on 1 of 1 shared result
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
- Training-Free Class Purification for Open-Vocabulary Semantic Segmentation (ICCV 2025 · similarity 0.91)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- Unveiling the Knowledge of CLIP for Training-Free Open-Vocabulary Semantic Segmentation (AAAI 2025 · similarity 0.90)
- OPMapper: Enhancing Open-Vocabulary Semantic Segmentation with Multi-Guidance Information (NeurIPS 2025 · similarity 0.90)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.88)
- Training-Free Open-Vocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation (CVPR 2024 · similarity 0.88)
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.
