Research map / Detection and segmentation / Open-set and few-shot detection
Hierarchical Open-vocabulary Universal Image 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 |
|---|---|---|---|
| decouple thing and stuff decoders instead of a unified decoder (no fusion) (ablation) | HIPIE · COCO panoptic | PQ: 45.1 -> 50.6 | Helped |
| decouple thing and stuff decoders instead of a unified decoder (no fusion) (ablation) | HIPIE · COCO instance | AP mask: 42.9 -> 43.6 | Helped |
| add early text-image fusion to decoupled model (things and stuff) (ablation) | HIPIE decoupled · RefCOCO val | oIoU: 67.6 -> 77.1 | Helped |
| use early fusion for stuff classes in addition to things (ablation) | HIPIE decoupled · COCO panoptic | PQ: 51.3 -> 50.0 | Hurt |
| early fusion for things only (late fusion for stuff) vs no fusion (ablation) | HIPIE decoupled · COCO panoptic | PQ: 50.6 -> 51.3 | Helped |
| add early fusion with a unified decoder (ablation) | unified decoder · COCO panoptic | PQ: 45.1 -> 44.6 | Hurt |
| combine classification logits with CLIP features at zero-shot inference (ablation) | HIPIE ViT-H · ADE20K A-150 | PQ: 18.1 -> 22.9 | Helped |
| combine classification logits with CLIP features at zero-shot inference (ablation) | HIPIE ViT-H · A-150 semantic | mIoU: 19.8 -> 29.0 | Helped |
| HIPIE ViT-H vs ODISE on zero-shot semantic segmentation | HIPIE ViT-H · PAS-21 / A-150 / PC-59 | mIoU: 29.9 -> 29.0 (A-150); 57.3 -> 59.3 (PC-59); 84.6 -> 83.3 (PAS-21) | Mixed |
Closest papers by meaning
- USE: Universal Segment Embeddings for Open-Vocabulary Image Segmentation (CVPR 2024 · similarity 0.91)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.90)
- SED: A Simple Encoder-Decoder for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Effective SAM Combination for Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.89)
- LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation (NeurIPS 2025 · similarity 0.88)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 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.