Research map / Video / Video segmentation and tracking
MeViS: A Multi-Modal Dataset for Referring Motion Expression Video Segmentation
Henghui Ding, Chang Liu, Shuting He, Kaining Ying, Xudong Jiang, Chen Change Loy, Yu–Gang Jiang
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 a large language model for motion reasoning (ablation) | LMPM · MeViSv2 | J&F: 38.3 -> 42.4 | Improved |
| add a large language model for motion reasoning (ablation) | LMPM · MeViSv1 | J&F: 42.2 -> 45.9 | Improved |
| add temporal-level contrastive loss (ablation) | LMPM+LLM · MeViSv2 | J&F: 42.4 -> 43.9 | Improved |
| add temporal-level contrastive loss (ablation) | LMPM+LLM · MeViSv1 | J&F: 45.9 -> 47.6 | Improved |
| add temporal context attention to image-based VLT (ablation) | VLT · MeViS | J&F: 27.8 -> 35.5 | Improved |
| add temporal context attention to ReferFormer (ablation) | ReferFormer · MeViS | J&F: 31.0 -> 36.3 | Improved |
| add temporal context to VLT (ablation) | VLT · Refer-YouTube-VOS | J&F: 63.1 -> 62.7 | No clear effect |
| add temporal context to VLT (ablation) | VLT · DAVIS17-RVOS | J&F: 60.4 -> 60.3 | No clear effect |
| train on referring image datasets instead of video data | VLT · MeViS | J&F: 27.8 -> 22.5 | Hurt |
| use adaptive multi-object output instead of top-1 mask | DsHmp/LMPM++ · MeViSv2 | N-acc.: 0.0 -> 21.0 | Improved |
| add detection head for direct box prediction | LMPM++ · MeViSv2 RMOT | HOTA*: 38.1 -> 38.8 | Improved |
| use ground-truth text vs audio input | LMPM · MeViSv2 AVOS | J&F: 38.3 -> 37.5 | Mixed |
Closest papers by meaning
- MeViS: A Large-scale Benchmark for Video Segmentation with Motion Expressions (ICCV 2023 · similarity 0.96)
- Decoupled Motion Expression Video Segmentation (CVPR 2025 · similarity 0.91)
- Decoupling Static and Hierarchical Motion Perception for Referring Video Segmentation (CVPR 2024 · similarity 0.90)
- Long-RVOS: A Comprehensive Benchmark for Long-term Referring Video Object Segmentation (CVPR 2026 · similarity 0.89)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.88)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.88)
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.
