Research map / Vision-language and multimodal / CLIP and zero-shot
Decoupling Zero-Shot Semantic Segmentation
Check 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 |
|---|---|---|---|
| decouple class-agnostic pixel grouping from segment-level zero-shot classification | ZS3 compared with pixel-level SPNet-FPN · COCO-Stuff (GZS3) | mIoU unseen: 11.0 -> 21.4 | Helped |
| use segment-level visual embeddings with CLIP text embeddings instead of ft+w2v | ZegFormer-seg · COCO-Stuff (GZS3) | mIoU unseen: 10.8 -> 21.4 | Helped |
| use segment-level classification rather than pixel-level classification | ZS3 models · COCO-Stuff (GZS3) | boundary F-score (F_b): 42.6 -> 50.4 | Helped |
| add the segment image-embedding classification branch and fuse its scores (ablation) | ZegFormer-seg · COCO-Stuff (GZS3) | mIoU unseen: 21.4 -> 33.1 | Helped |
| add the segment image-embedding classification branch and fuse its scores (ablation) | ZegFormer-seg · ADE20k-Full (GZS3) | mIoU unseen: 1.3 -> 5.6 | Helped |
| preprocess segment images by cropping and masking rather than cropping alone (ablation) | ZegFormer · COCO-Stuff (GZS3) | mIoU harmonic: 25.6 -> 34.4 | Helped |
| use masked segment images rather than crop-only segment images (ablation) | ZegFormer · COCO-Stuff (GZS3) | mIoU harmonic: 25.6 -> 33.3 | Helped |
| add prompt-template ensembling instead of using one template (ablation) | ZegFormer · COCO-Stuff (GZS3) | mIoU unseen: 32.7 -> 33.1 | Helped |
| use crop-only image preprocessing for segment classification (ablation) | ZegFormer · COCO-Stuff (GZS3) | mIoU harmonic: 27.2 -> 25.6 | Hurt |
Closest papers by meaning
- S3A: Towards Realistic Zero-Shot Classification via Self Structural Semantic Alignment (AAAI 2024 · similarity 0.90)
- A Simple Framework for Open-Vocabulary Zero-Shot Segmentation (ICLR 2025 · similarity 0.89)
- Cascade-CLIP: Cascaded Vision-Language Embeddings Alignment for Zero-Shot Semantic Segmentation (ICML 2024 · similarity 0.89)
- Label Propagation for Zero-shot Classification with Vision-Language Models (CVPR 2024 · similarity 0.88)
- Label-Efficient Few-Shot Semantic Segmentation with Unsupervised Meta-Training (AAAI 2024 · similarity 0.88)
- Exploring Regional Clues in CLIP for Zero-Shot Semantic Segmentation (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.