Research map / Video / Video segmentation and tracking
MPG-SAM 2: Adapting SAM 2 with Mask Priors and Global Context for Referring Video Object Segmentation
Fu Rong, Meng Lan, Qian Zhang, Lefei Zhang
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 mask-text similarity loss (ablation) | baseline SAM 2 + multimodal encoder · Ref-YouTube-VOS val | J&F: 69.4 -> 70.3 | Improved |
| Add mask prior generator (ablation) | baseline + L_sim · Ref-YouTube-VOS val | J&F: 70.3 -> 71.9 | Improved |
| Add hierarchical global-historical aggregator (ablation) | baseline + L_sim · Ref-YouTube-VOS val | J&F: 70.3 -> 72.3 | Improved |
| Add both MPG and HGA (ablation) | baseline + L_sim · Ref-YouTube-VOS val | J&F: 70.3 -> 73.9 | Improved |
| Use per-frame prompts instead of first-frame only (ablation) | baseline · Ref-YouTube-VOS val | J&F: 59.0 -> 69.4 | Improved |
| Add SAM 2 memory mechanism (ablation) | baseline · Ref-YouTube-VOS val | J&F: 66.2 -> 69.4 | Improved |
| Add spatiotemporal self-interaction in MPG (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.3 -> 73.9 | Improved |
| Add spatiotemporal cross-modal interaction in MPG (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.5 -> 73.9 | Improved |
| Use patch size 2 instead of 1 or 4 in HGA (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.2 -> 73.9 | Improved |
| Use masked global video feature instead of vanilla (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.1 -> 73.9 | Improved |
| Increase pixel-level fusion layers from 1 to 3 (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.9 -> 73.0 | Hurt |
| Increase object-level fusion layers from 1 to 3 (ablation) | MPG-SAM 2 · Ref-YouTube-VOS val | J&F: 73.9 -> 72.9 | Hurt |
Later papers that compared with this method
2 later accepted papers report MPG-SAM 2 as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Where It Moves, It Matters: Referring Surgical Instrument Segmentation via Motion (AAAI 2026 · better on 3 of 3 shared results)
- VIRST: Video-Instructed Reasoning Assistant for SpatioTemporal Segmentation (CVPR 2026 · better on 2 of 2 shared results)
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.
- VISA — VISA: Reasoning Video Object Segmentation via Large Language Model (ECCV 2024) · 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
- SAMWISE: Infusing Wisdom in SAM2 for Text-Driven Video Segmentation (CVPR 2025 · similarity 0.91)
- End-to-End Referring Video Object Segmentation With Multimodal Transformers (CVPR 2022 · similarity 0.90)
- SAM-I2V: Upgrading SAM to Support Promptable Video Segmentation with Less than 0.2% Training Cost (CVPR 2025 · similarity 0.90)
- SAM 2: Segment Anything in Images and Videos (ICLR 2025 · similarity 0.89)
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.89)
- Semantic and Sequential Alignment for Referring Video Object Segmentation (CVPR 2025 · similarity 0.89)
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.
