Research map / Detection and segmentation / Open-set and few-shot detection
Primitive Generation and Semantic-Related Alignment for Universal Zero-Shot Segmentation
Shuting He, Henghui Ding, Wei Jiang
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 |
|---|---|---|---|
| Replace the GMMN generator with a primitive generator (ablation) | Zero-shot panoptic segmentation model · MSCOCO zero-shot panoptic segmentation | HM PQ: at least 9.0% PQ | Improved |
| Add semantic relationship alignment (ablation) | Primitive generator · MSCOCO zero-shot panoptic segmentation | HM PQ: 2.6% | Improved |
| Add feature disentanglement to alignment (ablation) | Primitive generator with alignment · MSCOCO zero-shot panoptic segmentation | HM PQ: 2.0% HM-PQ | Improved |
| Add feature disentanglement to alignment (ablation) | Primitive generator with alignment · MSCOCO zero-shot panoptic segmentation | HM SQ: 13.3% HM-SQ | Improved |
| Add feature disentanglement to alignment (ablation) | Primitive generator with alignment · MSCOCO zero-shot panoptic segmentation | HM RQ: 2.7% HM-RQ | Improved |
| Increase primitives from 100 to 400 (ablation) | Primitive generator · MSCOCO zero-shot panoptic segmentation | PQ: 4.2% | Improved |
| Increase primitives beyond 400 (ablation) | Primitive generator · MSCOCO zero-shot panoptic segmentation | PQ | Hurt |
| Use primitive generation, relationship alignment, and disentanglement instead of a GMMN generator (ablation) | Zero-shot panoptic segmentation model · MSCOCO zero-shot panoptic segmentation | HM PQ: 8.7 -> 22.3 | Improved |
| Use primitive generation, relationship alignment, and disentanglement instead of a projection-based classifier (ablation) | Zero-shot panoptic segmentation model · MSCOCO zero-shot panoptic segmentation | unseen PQ: 0.0 -> 15.3 | Improved |
| Use PADing instead of ZegFormer-seg | Zero-shot semantic segmentation · COCO-Stuff | HM IoU: 3.5% HM-IoU | Improved |
| Use PADing instead of ZSI | Zero-shot instance segmentation · GZSI 48/17 | HM mAP: 7.20% HM-mAP | Improved |
| Use PADing instead of ZSI | Zero-shot instance segmentation · GZSI 48/17 | HM Recall: 5.27% HM-Recall | Improved |
Closest papers by meaning
- Semantic-Centric Alignment for Zero-shot Panoptic Segmentation with Limited Data (IJCV 2026 · similarity 0.92)
- Consistent Structural Relation Learning for Zero-Shot Segmentation (NeurIPS 2020 · similarity 0.92)
- Exploiting a Joint Embedding Space for Generalized Zero-Shot Semantic Segmentation (ICCV 2021 · similarity 0.92)
- Zero-Shot Semantic Segmentation (NeurIPS 2019 · similarity 0.91)
- Prototypical Matching and Open Set Rejection for Zero-Shot Semantic Segmentation (ICCV 2021 · similarity 0.91)
- Class Semantic Attribute Perception Guided Zero-Shot Learning (AAAI 2025 · 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.
