Research map / Detection and segmentation / CNN features
SegGPT: Towards Segmenting Everything in Context
Xinlong Wang, Xiaosong Zhang, Yue Cao, Wen Wang, Chunhua Shen, Tiejun Huang
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 feature ensemble instead of spatial ensemble for multiple examples (ablation) | SegGPT · DAVIS 2017 | J&F: 61.9 -> 74.7 | Improved |
| use feature ensemble instead of spatial ensemble on low-resolution data (ablation) | SegGPT · FSS-1000 | mIoU: 89.3 -> 87.8 | Hurt |
| use 4 example frames with feature ensemble instead of 1 (ablation) | SegGPT · DAVIS 2017 | J&F: 70.0 -> 74.7 | Improved |
| increase context frames from 8 to 16 (ablation) | SegGPT · DAVIS 2017 | J&F: 75.6 -> 74.6 | Hurt |
| use random color mapping in in-context training instead of fixed task colors | SegGPT · ADE20K semantic segmentation | mIoU: Painter 49.9 -> SegGPT 39.6 | Hurt |
| use random color mapping in in-context training instead of fixed task colors | SegGPT · COCO panoptic segmentation | PQ: Painter 43.4 -> SegGPT 34.4 | Hurt |
| use random coloring and mixed training instead of Painter's fixed colors | SegGPT · FSS-1000 one-shot (out-of-domain) | mIoU: 61.7 -> 85.6 | Improved |
| use random coloring and mixed training instead of Painter | SegGPT · YouTube-VOS 2018 | G: 24.1 -> 74.7 | Improved |
| use SegGPT versus specialist VOS models trained on video data | SegGPT · YouTube-VOS 2018 | G: SegGPT 74.7 vs XMem 86.1, AGAME 66.0 | Mixed |
Later papers that compared with this method
12 later accepted papers report SegGPT as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- CDICS: Delving Into Fine-Grained Attribute for In-Context Segmentation via Compositional Prompts and Phased Decoupling (CVPR 2026 · better on 3 of 4 shared results)
- FoundYou: A Unified Model for Personalized Segmentation and Retrieval (ECCV 2026 · better on 4 of 4 shared results)
- Medverse: A Universal Model for Full-Resolution 3D Medical Image Segmentation, Transformation and Enhancement (AAAI 2026 · better on 1 of 1 shared result)
- MedSAMix: A Training-Free Model Merging Approach for Medical Image Segmentation (AAAI 2026 · better on 1 of 1 shared result)
- One Polyp Identifies All: One-Shot Polyp Segmentation with SAM via Cascaded Priors and Iterative Prompt Evolution (ICCV 2025 · better on 3 of 3 shared results)
- Unified Medical Lesion Segmentation via Self-referring Indicator (CVPR 2025 · better on 1 of 1 shared result)
- Neuroverse3D: Developing In-Context Learning Universal Model for Neuroimaging in 3D (ICCV 2025 · better on 1 of 1 shared result)
- A Simple Image Segmentation Framework via In-Context Examples (NeurIPS 2024 · better on 2 of 4 shared results)
4 more not listed.
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
- Seg-ReSearch: Segmentation with Interleaved Reasoning and External Search (ICML 2026 · similarity 0.89)
- A Simple Image Segmentation Framework via In-Context Examples (NeurIPS 2024 · similarity 0.88)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.88)
- OMG-Seg: Is One Model Good Enough For All Segmentation? (CVPR 2024 · similarity 0.87)
- Explore In-Context Segmentation via Latent Diffusion Models (AAAI 2025 · similarity 0.87)
- Uncertainty-aware Fine-tuning of Segmentation Foundation Models (NeurIPS 2024 · 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.
