Research map / Detection and segmentation / Open-set and few-shot detection
Unified Open-World Segmentation with Multi-Modal Prompts
Yang Liu, Yufei Yin, Chenchen Jing, Muzhi Zhu, Hao Chen, Yuling Xi, Bo Feng, Hao Wang, Shiyu Li, Chunhua Shen
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 text-branch training to vision branch (ablation) | COSINE (10k steps) · LVIS-92i 1-shot | mIoU: 24.5 -> 27.7 | Improved |
| add vision-branch training to text branch (ablation) | COSINE (10k steps) · ADE20K | PQ: 13.2 -> 17.7 | Improved |
| add text prompt at inference (ablation) | COSINE · LVIS-92i 1-shot | mIoU: 35.2 -> 43.1 | Improved |
| add visual prompt at inference (ablation) | COSINE · ADE20K | PQ: 31.0 -> 31.4 | Improved |
| add CLIP encoder to DINOv2 pool (ablation) | COSINE (10k steps) · ADE20K | PQ: 9.7 -> 17.7 | Improved |
| add DINOv2 encoder to CLIP pool (ablation) | COSINE (10k steps) · ADE20K | PQ: 11.4 -> 17.7 | Improved |
| add CLIP encoder to DINOv2 pool (ablation) | COSINE (10k steps) · LVIS-92i 1-shot | mIoU: 27.7 -> 27.7 | No clear effect |
| add Feature Blending (ablation) | COSINE (10k steps) · ADE20K | PQ: 9.4 -> 17.7 | Improved |
| add Image-Prompt Aligner (ablation) | COSINE (10k steps) · ADE20K | PQ: 12.0 -> 17.7 | Improved |
| add prompt refining in Multi-Modality Decoder (ablation) | COSINE (10k steps) · ADE20K | PQ: 16.5 -> 17.7 | Improved |
| add prompt refining (ablation) | COSINE (10k steps) · ADE20K | mIoU: 30.9 -> 30.4 | Hurt |
| fine-tune on pure image prompt dataset (COSINE-FT) | COSINE · DAVIS 2017 | J&F: 76.7 -> 80.2 | Improved |
Closest papers by meaning
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.90)
- COS3D: Collaborative Open-Vocabulary 3D Segmentation (NeurIPS 2025 · similarity 0.89)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.89)
- Emergent Open-Vocabulary Semantic Segmentation from Off-the-shelf Vision-Language Models (CVPR 2024 · similarity 0.88)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.88)
- DPSeg: Dual-Prompt Cost Volume Learning for Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.87)
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.
