Research map / Detection and segmentation / Open-set and few-shot detection
CURE: Contextual Debiasing and Unbiased Refinement for Training-Free Open-Vocabulary Semantic Segmentation
Jiean Wang, Sanqing Qu, Fan Lu, Huanhuan Bao, Wei Tian, Bo Jin, Jiangtong Li, Junqiao Zhao, Guang Chen
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 Reject and Reconstruct (RR) (ablation) | ClearCLIP · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 28.2 -> 28.9 | Improved |
| add Global Contextual Debiasing (GCD) (ablation) | ClearCLIP · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 28.2 -> 29.1 | Improved |
| add Spatial Correlation Refinement (SCR) (ablation) | ClearCLIP · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 28.2 -> 29.4 | Improved |
| add SCR on top of RR and GCD (ablation) | CURE (RR+GCD) · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 29.5 -> 30.9 | Improved |
| add RR on top of GCD and SCR (ablation) | CURE (GCD+SCR) · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 30.6 -> 30.9 | Improved |
| replace direct CLS subtraction with GCD (ablation) | CURE · ADE20K/PC59/Cityscapes avg (ViT-B/16) | mIoU (%): 29.8 -> 30.9 | Improved |
| add CURE to ClearCLIP | ClearCLIP (ViT-B/16) · 5-dataset average (VOC20, PC59, ADE20K, Cityscapes, COCO-Stuff) | mIoU (%): 37.9 -> 40.1 | Improved |
| add CURE to ClearCLIP | ClearCLIP (ViT-L/14) · 5-dataset average (VOC20, PC59, ADE20K, Cityscapes, COCO-Stuff) | mIoU (%): 35.3 -> 39.6 | Improved |
| add CURE to SCLIP | SCLIP (ViT-L/14) · 5-dataset average (VOC20, PC59, ADE20K, Cityscapes, COCO-Stuff) | mIoU (%): 24.4 -> 40.1 | Improved |
| add CURE to SCCLIP | SCCLIP (ViT-B/16) · 5-dataset average (VOC20, PC59, ADE20K, Cityscapes, COCO-Stuff) | mIoU (%): 42.4 -> 42.9 | Improved |
| add CURE to ClearCLIP with MetaCLIP pretraining | ClearCLIP · ADE20K/COCO/PC59 avg | mIoU (%): 25.2 -> 27.6 | Improved |
| add CURE to ClearCLIP on long-tail classes | ClearCLIP · Cityscapes long-tail | mIoU (%): 26.4 -> 27.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ClearCLIP — ClearCLIP: Decomposing CLIP Representations for Dense Vision-Language Inference (ECCV 2024) · better on 6 of 7 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
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- Open-Vocabulary Segmentation with Semantic-Assisted Calibration (CVPR 2024 · similarity 0.89)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.89)
- Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.89)
- Training-Free Open-Vocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation (CVPR 2024 · similarity 0.89)
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.
