Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Pre-Training With Language-Vision Prompts for Low-Data Instance Segmentation
Dingwen Zhang, Hao Li, Diqi He, Nian Liu, Lechao Cheng, Jingdong Wang, Junwei Han
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 auxiliary kernel supervision loss (ablation) | UPLVP K-Net · COCO 10% | mask AP: 25.3 -> 25.4 | Improved |
| add cosine-similarity prompt-kernel matching vs no prompting (ablation) | UPLVP K-Net · COCO 10% | mask AP: 25.2 -> 25.4 | Improved |
| random prompt assignment instead of no prompt (ablation) | UPLVP K-Net · COCO 10% | mask AP: -0.1 mAP | Hurt |
| cosine similarity instead of sequential assignment (ablation) | UPLVP K-Net · COCO 10% | mask AP: 25.3 -> 25.4 | Improved |
| use 80 CLIP category labels instead of saliency 0/1 labels (ablation) | UPLVP K-Net · COCO 10% | mask AP: 24.2 -> 25.4 | Improved |
| add prompting to K-Net with UPSP pseudo labels (ablation) | K-Net · COCO 10% | mask AP: 21.7 -> 23.5 | Improved |
| add prompting to K-Net with random proposals (ablation) | K-Net · COCO 10% | mask AP: 0.8 -> 10.0 | Improved |
| add prompting with FreeSOLO pseudo labels (ablation) | K-Net · COCO 10% | mask AP: 23.3 -> 24.0 | Improved |
| UPLVP pre-training instead of ImageNet supervised | K-Net · COCO 5% | mask AP: 14.8 -> 22.2 | Improved |
| UPLVP instead of UPSP (add category knowledge) | K-Net · COCO 5% | mask AP: 19.9 -> 22.2 | Improved |
| UPLVP vs FreeSOLO pre-training | K-Net · COCO 10% | mask AP: 25.6 -> 25.4 | No clear effect |
| UPLVP vs Mask R-CNN on small objects | K-Net · Cityscapes | AP_S: 8.3 -> 7.3 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- K-Net — K-Net: Towards Unified Image Segmentation (NeurIPS 2021) · better on 2 of 2 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
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.87)
- Prompt-Driven Referring Image Segmentation with Instance Contrasting (CVPR 2024 · similarity 0.86)
- Locality Alignment Improves Vision-Language Models (ICLR 2025 · similarity 0.85)
- Training-Free Open-Ended Object Detection and Segmentation via Attention as Prompts (NeurIPS 2024 · similarity 0.85)
- Prompting Diffusion Models for Zero-Shot Instance Segmentation (NeurIPS 2026 · similarity 0.85)
- PromptKD: Unsupervised Prompt Distillation for Vision-Language Models (CVPR 2024 · similarity 0.85)
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.
