Research map / 3D vision / 3D generation
V^2-SAM: Marrying SAM2 with Multi-Prompt Experts for Cross-View Object Correspondence
Jiancheng Pan, Runze Wang, Tianwen Qian, Mohammad Mahdi, Yanwei Fu, Xiangyang Xue, Xiaomeng Huang, Luc Van Gool, Danda Pani Paudel, Yuqian Fu
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 centroid-of-query-mask prompt with V2-Anchor DINOv3 correspondence prompt (ablation) | Anchor Expert · Ego-Exo4D v2 test | Total IoU: 1.5 -> 40.1 | Improved |
| add Visual Prompt Matcher instead of raw mask-pooled query feature (ablation) | Visual Expert · Ego-Exo4D v2 test | Total IoU: 3.0 -> 41.4 | Improved |
| combine Anchor and Visual experts (ablation) | V2-SAM · Ego-Exo4D v2 test (Ego2Exo) | IoU: 38.7/36.2 -> 42.7 | Improved |
| add Fusion Expert and PCCS selection to A+B experts (ablation) | V2-SAM · Ego-Exo4D v2 test | Total IoU: 45.5 -> 48.0 | Improved |
| use multi-experts with PCCS instead of single Fusion Expert (ablation) | V2-SAM · HANDAL-X (zero-shot) | IoU: 66.4 -> 77.2 | Improved |
| use multi-experts instead of single expert (ablation) | V2-SAM · Ego-Exo4D v2 test | Total IoU: 45.9 -> 48.0 | Improved |
| multi-expert variant versus single-expert on localization error (ablation) | V2-SAM · Ego-Exo4D v2 test (Ego2Exo) | Loc.E (lower better): 0.055 -> 0.056 | No clear effect |
| Fusion Expert alone vs two-expert combination (ablation) | V2-SAM · Ego-Exo4D v2 test | Total IoU: 45.9 vs 45.5 | No clear effect |
Closest papers by meaning
- ObjectRelator: Enabling Cross-View Object Relation Understanding Across Ego-Centric and Exo-Centric Perspectives (ICCV 2025 · similarity 0.91)
- Robust Ego-Exo Correspondence with Long-Term Memory (NeurIPS 2025 · similarity 0.88)
- Self-Supervised Cross-View Correspondence with Predictive Cycle Consistency (CVPR 2025 · similarity 0.86)
- Cross-View Completion Models are Zero-shot Correspondence Estimators (CVPR 2025 · similarity 0.85)
- C3Po: Cross-View Cross-Modality Correspondence by Pointmap Prediction (NeurIPS 2025 · similarity 0.85)
- Understanding Multi-View Transformers (NeurIPS 2026 · similarity 0.84)
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.
