Research map / Vision-language and multimodal / CLIP and zero-shot
Fine-Grained Visual Text Prompting
Lingfeng Yang, Xiang Li, Yueze Wang, Xinlong Wang, Jian Jun Yang
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 |
|---|---|---|---|
| replace colored mask by blur reverse mask (ablation) | FGVP · RefCOCO REC | accuracy: 41.8 -> 52.8 | Improved |
| blur background instead of graying it (ablation) | FGVP · RefCOCO REC | accuracy: 45.8 -> 52.8 | Improved |
| use precise semantic mask instead of box for blur reverse (ablation) | FGVP · RefCOCO REC | accuracy: 48.8 -> 52.8 | Improved |
| use blur reverse mask instead of red circle prompt | FGVP vs RedCircle · RefCOCO val (UNINEXT proposals) | accuracy: 48.9 -> 52.8 | Improved |
| use blur reverse mask instead of red circle prompt | FGVP vs RedCircle · RefCOCO val (CPT codebase) | accuracy: 38.0 -> 46.1 | Improved |
| use FGVP prompts in ReCLIP ensemble with relations | FGVP vs RedCircle · RefCOCO val (ReCLIP codebase) | accuracy: 56.8 -> 59.6 | Improved |
| use blur reverse mask instead of crop for COCO object classification (ablation) | FGVP · COCO ground-truth prompts | accuracy: 70.9 -> 67.8 | Hurt |
| use SAM huge instead of base (ablation) | FGVP · RefCOCO val | accuracy: 51.7 -> 51.9 | Improved |
| enable mask-filter post-processing (SAM huge) (ablation) | FGVP · RefCOCO val | accuracy: 51.9 -> 52.8 | Improved |
| enable mask-filter post-processing (SAM base) (ablation) | FGVP · RefCOCO val | accuracy: 51.7 -> 52.1 | Improved |
| set Gaussian blur sigma to 100 (ablation) | FGVP · 8 benchmarks | accuracy | Improved |
| expand or shrink mask scale away from 1 (ablation) | FGVP · mask preciseness | accuracy | Hurt |
Closest papers by meaning
- Visual In-Context Prompting (CVPR 2024 · similarity 0.91)
- Just a Few Glances: Open-Set Visual Perception with Image Prompt Paradigm (AAAI 2025 · similarity 0.87)
- SA²VP: Spatially Aligned-and-Adapted Visual Prompt (AAAI 2024 · similarity 0.87)
- Draw-and-Understand: Leveraging Visual Prompts to Enable MLLMs to Comprehend What You Want (ICLR 2025 · similarity 0.87)
- Hugging Visual Prompt and Segmentation Tokens: Consistency Learning for Fine-Grained Visual Understanding in MLLMs (CVPR 2026 · similarity 0.87)
- STOP: Integrated Spatial-Temporal Dynamic Prompting for Video Understanding (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.
