Research map / Detection and segmentation / Open-set and few-shot detection
Rethinking and Improving Visual Prompt Selection for In-Context Learning Segmentation Framework
Wei Suo, Lanqing Lai, Mengyang Sun, Hanwang Zhang, Peng Wang, Yanning Zhang
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 nearest-and-farthest sampling per cluster to the candidate pool (ablation) | SCS · COCO-20i (5-shot) | mIoU: 55.7 -> 57.6 | Improved |
| add adaptive search module (RL search agent) (ablation) | SCS · COCO-20i (5-shot) | mIoU: 57.6 -> 59.5 | Improved |
| compact candidate pool by clustering and using centroid-nearest examples (ablation) | SegGPT prompt selection · COCO-20i (5-shot) | mIoU: 56.1 -> 55.7 | No clear effect |
| replace random prompts with SCS prompts | SegGPT · COCO-20i (1-shot) | mIoU: 48.6 -> 57.2 | Improved |
| replace similarity sorting with SCS | MAE-VQGAN · PASCAL-5i (1-shot) | mIoU: 33.6 -> 35.0 | Improved |
| randomly sample in-context examples with different seeds | SegGPT · PASCAL-5i | mIoU: best minus worst 5.6 (1-shot), 4.1 (5-shot) | Mixed |
| combine nearest and farthest examples versus two nearest (ablation) | SegGPT with ViT similarity · PASCAL | mIoU: +2.4 mIoU | Improved |
| swap CLIP-ViT encoder for ViT or CLIP-resnet (ablation) | SCS · COCO-20i (5-shot) | mIoU: 59.5 -> 58.7 / 58.5 | No clear effect |
| use density sorting instead of nearest-farthest sampling (ablation) | SCS · COCO-20i (5-shot) | mIoU: 59.5 -> 59.1 | No clear effect |
Closest papers by meaning
- Visual In-Context Prompting (CVPR 2024 · similarity 0.90)
- Towards Global Optimal Visual In-Context Learning Prompt Selection (NeurIPS 2024 · similarity 0.89)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.88)
- A Simple Image Segmentation Framework via In-Context Examples (NeurIPS 2024 · similarity 0.88)
- Rethinking Interactive Image Segmentation with Low Latency High Quality and Diverse Prompts (CVPR 2024 · similarity 0.88)
- Visual Prompting for Generalized Few-shot Segmentation: A Multi-scale Approach (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.
