Research map / Detection and segmentation / Open-set and few-shot detection
Prototypical Matching and Open Set Rejection for Zero-Shot Semantic Segmentation
Hui Zhang, Henghui Ding
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 open-set rejection (ablation) | prototype matching ZSS · Pascal VOC 2012 (cow/motorbike unseen) | hIoU: 5.4 -> 60.5 | Improved |
| add open-set rejection (ablation) | prototype matching ZSS · Pascal VOC 2012 (cat/tv unseen) | hIoU: 6.1 -> 62.2 | Improved |
| add open-set rejection (ablation) | prototype matching ZSS · Pascal VOC 2012 (cow/motorbike unseen) | unseen mIoU: 2.8 -> 51.3 | Improved |
| use ground-truth rejection instead of learned open-set rejection (oracle upper bound) (ablation) | open-set rejection · Pascal VOC 2012 (cow/motorbike unseen) | hIoU: 60.5 -> 69.8 | Improved |
| use ground-truth rejection instead of learned open-set rejection (oracle upper bound) (ablation) | open-set rejection · Pascal VOC 2012 (cat/tv unseen) | hIoU: 62.2 -> 70.1 | Improved |
| replace direct word-embedding prototypes with a learned projection network (ablation) | prototype matching ZSS · Pascal VOC 2012 (aeroplane/sofa unseen) | hIoU: 65.0 -> 70.2 | Improved |
| replace direct word-embedding prototypes with a learned projection network (ablation) | prototype matching ZSS · Pascal VOC 2012 (bird/boat unseen) | hIoU: 62.7 -> 66.9 | Improved |
| use trainable prototypes instead of a convolutional classifier (ablation) | supervised segmentation network · Pascal VOC 2012 (supervised) | mIoU: 76.8 -> 76.9 | No clear effect |
| proposed method versus CSRL | prototypical matching + OSR · Pascal VOC 2012 (2 unseen classes) | hIoU: 56.3 -> 60.5 | Improved |
| proposed method versus CSRL as unseen classes grow to 10 | prototypical matching + OSR · Pascal VOC 2012 (10 unseen classes) | hIoU: 31.0 -> 46.1 | Improved |
| proposed method versus ZS3+GC | prototypical matching + OSR · Pascal Context (2 unseen classes) | hIoU: 34.8 -> 46.3 | Improved |
| proposed method versus CaGNet | prototypical matching + OSR · Pascal Context (10 unseen classes) | hIoU: 21.2 -> 27.7 | Improved |
Closest papers by meaning
- Primitive Generation and Semantic-Related Alignment for Universal Zero-Shot Segmentation (CVPR 2023 · similarity 0.91)
- Training-free Open-Vocabulary Semantic Segmentation via Diverse Prototype Construction and Sub-region Matching (AAAI 2025 · similarity 0.90)
- Zero-Shot Semantic Segmentation (NeurIPS 2019 · similarity 0.90)
- Delving Into Shape-Aware Zero-Shot Semantic Segmentation (CVPR 2023 · similarity 0.90)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.88)
- CSL: Class-Agnostic Structure-Constrained Learning for Segmentation Including the Unseen (AAAI 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.
