Research map / Learning theory and algorithms / Optimal transport
Click Prompt Learning with Optimal Transport for Interactive Segmentation
Jie Liu, Haochen wang, Wenzhe Yin, Jan-Jakob Sonke, Efstratios Gavves
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 the prompt-pixel alignment module (PPAM) alone to the SimpleClick baseline (ablation) | SimpleClick ViT-B (SBD-trained) · DAVIS | NoC90: 5.48 -> 5.34 | Improved |
| add the click prompt optimal transport module (CPOT) alone to the baseline (ablation) | SimpleClick ViT-B (SBD-trained) · Berkeley | NoC90: 2.46 -> 2.32 | Improved |
| combine PPAM and CPOT versus either module alone (ablation) | CPlot ViT-B (SBD-trained) · Berkeley | NoC90: 2.37 -> 2.18 (vs PPAM only) | Improved |
| add PPAM alone on GrabCut (ablation) | SimpleClick ViT-B (SBD-trained) · GrabCut | NoC90: 1.54 -> 1.54 | No clear effect |
| initialize click prompts with position and content prompts in addition to the learnable prompt (ablation) | CPlot · DAVIS | NoC90: 5.89 -> 5.29 | Improved |
| use five click prompts instead of one (ablation) | CPlot · Berkeley | NoC90: 2.69 -> 2.18 | Improved |
| increase click prompts from five to seven (ablation) | CPlot · Berkeley | NoC90: 2.18 -> 2.46 | Hurt |
| increase PPAM transformer blocks from two to three or four (ablation) | CPlot · DAVIS | NoC90: 5.29 -> 5.69 (L=3) | Hurt |
| apply CPlot instead of SimpleClick when trained on COCO+LVIS | ViT-B · DAVIS | NoC85: 3.66 -> 3.26 | Improved |
| apply CPlot to out-of-domain medical images without fine-tuning | ViT-B (COCO+LVIS) · OAIZIB | mIoU@10: 76.05 -> 78.45 | Improved |
| CPlot versus PseudoClick on Pascal VOC with COCO+LVIS training | ViT-B · Pascal VOC | NoC90: 2.31 vs 2.25 (PseudoClick) | Mixed |
Closest papers by meaning
- RClicks: Realistic Click Simulation for Benchmarking Interactive Segmentation (NeurIPS 2024 · similarity 0.90)
- TETRIS: Towards Exploring the Robustness of Interactive Segmentation (AAAI 2024 · similarity 0.89)
- Rethinking Interactive Image Segmentation with Low Latency High Quality and Diverse Prompts (CVPR 2024 · similarity 0.88)
- NTClick: Achieving Precise Interactive Segmentation With Noise-tolerant Clicks (CVPR 2025 · similarity 0.88)
- Repurposing Stable Diffusion Attention for Training-Free Unsupervised Interactive Segmentation (CVPR 2025 · similarity 0.87)
- Order-aware Interactive Segmentation (ICLR 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.
