Research map / Detection and segmentation / Open-set and few-shot detection
FreeSeg: Unified, Universal and Open-Vocabulary Image Segmentation
Jie Qin, Jie Wu, Pengxiang Yan, Ming Li, Ren Yuxi, Xuefeng Xiao, Yitong Wang, Rui Wang, Shilei Wen, Xin Pan, Xingang Wang
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 |
|---|---|---|---|
| use FreeSeg instead of ZSSeg | open-vocabulary semantic segmentation · COCO | mIoU unseen: 43.6 -> 49.1 | Improved |
| use FreeSeg instead of ZSSeg | open-vocabulary semantic segmentation · ADE20K | mIoU unseen: 20.3 -> 28.6 | Improved |
| use FreeSeg instead of ZSI | open-vocabulary instance segmentation · COCO | mAP unseen: 13.6 -> 20.6 | Improved |
| use FreeSeg instead of ZSSeg | open-vocabulary panoptic segmentation · COCO | PQ unseen: 9.7 -> 29.8 | Improved |
| add an adaptive class prompt (ablation) | FreeSeg · COCO semantic segmentation | mIoU unseen: 4.9 -> 38.5 | Improved |
| add an adaptive task prompt (ablation) | FreeSeg with adaptive class prompt · COCO semantic segmentation | mIoU unseen: 38.5 -> 41.3 | Improved |
| add semantic context interaction (ablation) | FreeSeg with adaptive class and task prompts · COCO semantic segmentation | mIoU unseen: 41.3 -> 42.6 | Improved |
| add test-time prompt tuning (ablation) | FreeSeg with adaptive prompts and context interaction · COCO semantic segmentation | mIoU unseen: 42.6 -> 43.3 | Improved |
| use unified multi-task training instead of single-task training (ablation) | FreeSeg · COCO semantic segmentation | mIoU unseen: 42.9 -> 43.3 | Improved |
| use adaptive prompts instead of fixed templates (ablation) | FreeSeg · COCO semantic segmentation | mIoU unseen: 33.7 -> 43.3 | Improved |
Later papers that compared with this method
4 later accepted papers report FreeSeg as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Relationship Prompt Learning is Enough for Open-Vocabulary Semantic Segmentation (NeurIPS 2024 · better on 2 of 2 shared results)
- TagCLIP: Improving Discrimination Ability of Zero-Shot Semantic Segmentation (TPAMI 2024 · better on 1 of 1 shared result)
- Learning Mask-aware CLIP Representations for Zero-Shot Segmentation (NeurIPS 2023 · better on 9 of 9 shared results)
- Convolutions Die Hard: Open-Vocabulary Segmentation with Single Frozen Convolutional CLIP (NeurIPS 2023 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Exploring Simple Open-Vocabulary Semantic Segmentation (CVPR 2025 · similarity 0.90)
- OMG-Seg: Is One Model Good Enough For All Segmentation? (CVPR 2024 · similarity 0.89)
- APSeg: Auto-Prompt Network for Cross-Domain Few-Shot Semantic Segmentation (CVPR 2024 · similarity 0.89)
- CAT-Seg: Cost Aggregation for Open-Vocabulary Semantic Segmentation (CVPR 2024 · similarity 0.89)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.88)
- USE: Universal Segment Embeddings for Open-Vocabulary Image 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.
