Research map / Detection and segmentation / Open-set and few-shot detection
Uncovering Prototypical Knowledge for Weakly Open-Vocabulary Semantic Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 image-level prototypical regularization (I-NPR) (ablation) | PGSeg baseline · PASCAL VOC12 val | mIoU (%): 50.76 -> 52.51 | Helped |
| add text-level prototypical regularization (T-NPR) (ablation) | PGSeg baseline · PASCAL VOC12 val | mIoU (%): 50.76 -> 51.34 | Helped |
| add hard rejecting strategy with threshold 0.1 to I-NPR (ablation) | PGSeg · PASCAL VOC12 val | mIoU (%): 52.51 -> 52.92 | Helped |
| use HRS threshold 0.5 rather than I-NPR alone (ablation) | PGSeg · PASCAL VOC12 val | mIoU (%): 52.51 -> 51.62 | Hurt |
| use HRS threshold 0.3 rather than I-NPR alone (ablation) | PGSeg · PASCAL VOC12 val | mIoU (%): 52.51 -> 52.57 | Helped |
| combine I-NPR, T-NPR, and HRS with threshold 0.1 (ablation) | PGSeg · PASCAL VOC12 val | mIoU (%): 50.76 -> 53.24 | Helped |
| use 32 prototypes at level 1 instead of 64, with 8 at level 2 (ablation) | PGSeg without HRS · PASCAL VOC12 val | mIoU (%): 52.83 -> 52.49 | Hurt |
| use 128 prototypes at level 1 instead of 64, with 8 at level 2 (ablation) | PGSeg without HRS · PASCAL VOC12 val | mIoU (%): 52.83 -> 52.62 | Hurt |
| use 4 prototypes at level 2 instead of 8, with 64 at level 1 (ablation) | PGSeg without HRS · PASCAL VOC12 val | mIoU (%): 52.83 -> 52.71 | Hurt |
| use 16 prototypes at level 2 instead of 8, with 64 at level 1 (ablation) | PGSeg without HRS · PASCAL VOC12 val | mIoU (%): 52.83 -> 52.23 | Hurt |
| use PGSeg rather than GroupViT with the same 24M training-data volume | ViT-S segmentation models · PASCAL VOC12 val | mIoU (%): 50.8 -> 53.2 | Helped |
| use PGSeg rather than GroupViT with the same 24M training-data volume | ViT-S segmentation models · COCO val | mIoU (%): 27.5 -> 28.7 | Helped |
Closest papers by meaning
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.92)
- Exploring Simple Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.89)
- CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.88)
- USE: Universal Segment Embeddings for Open-Vocabulary Image Segmentation (CVPR 2024 · similarity 0.88)
- Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation (CVPR 2025 · 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.