Research map / Video / Video segmentation and tracking
Correspondence as Video: Test-Time Adaption on SAM2 for Reference Segmentation in the Wild
Haoran Wang, Zekun Li, Jian Zhang, Lei Qi, Yinghuan Shi
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 DBST diffusion transition frames (ablation) | SAM2 concat baseline · CD-FSS avg 1-shot | mIoU: 60.68 -> 62.68 | Improved |
| add TTGA test-time geometric alignment (ablation) | SAM2 + DBST · CD-FSS avg 1-shot | mIoU: 62.68 -> 64.06 | Improved |
| add DBST (ablation) | DEVA concat baseline · CD-FSS avg 1-shot | mIoU: 49.57 -> 57.31 | Improved |
| add TTGA (ablation) | DEVA + DBST · CD-FSS avg 1-shot | mIoU: 57.31 -> 59.93 | Improved |
| replace concatenation with mixup frames (ablation) | SAM2 concat baseline · CD-FSS avg 1-shot | mIoU: 60.68 -> 52.21 | Hurt |
| replace concatenation with affine augmentation frames (ablation) | SAM2 concat baseline · CD-FSS avg 1-shot | mIoU: 60.68 -> 56.84 | Hurt |
| use ACC instead of ABC as TTGA objective (ablation) | CAV-SAM · CD-FSS avg 1-shot | mIoU: 61.54 -> 64.06 | Improved |
| add TTGA with ABC objective (ablation) | SAM2 + DBST · CD-FSS avg 1-shot | mIoU: 62.08 -> 61.54 | Hurt |
| add CAV-SAM vs best prior average (APSeg) | SAM2-tiny · CD-FSS avg 1-shot | mIoU: 61.30 -> 64.06 | Improved |
| apply CAV-SAM on Deepglobe | SAM2-tiny · Deepglobe 1-shot | mIoU: 42.60 -> 39.11 | Hurt |
| apply CAV-SAM on Chest X-Ray | SAM2-tiny · Chest X-Ray 1-shot | mIoU: 84.10 -> 86.97 | Improved |
Closest papers by meaning
- SAM-I2V: Upgrading SAM to Support Promptable Video Segmentation with Less than 0.2% Training Cost (CVPR 2025 · similarity 0.92)
- SAM 2: Segment Anything in Images and Videos (ICLR 2025 · similarity 0.91)
- SAMWISE: Infusing Wisdom in SAM2 for Text-Driven Video Segmentation (CVPR 2025 · similarity 0.89)
- SAM2-LOVE: Segment Anything Model 2 in Language-aided Audio-Visual Scenes (CVPR 2025 · similarity 0.89)
- VidSeg: Training-free Video Semantic Segmentation based on Diffusion Models (CVPR 2025 · similarity 0.88)
- Personalize Segment Anything Model with One Shot (ICLR 2024 · 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.
