Research map / Video / Video-language understanding
End-to-End Referring Video Object Segmentation With Multimodal Transformers
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use MTTR instead of CMPC-V | RVOS · A2D-Sentences | mAP: 40.4 -> 44.7 | Helped |
| Use MTTR instead of CMPC-V | RVOS · JHMDB-Sentences | mAP: 34.2 -> 36.6 | Helped |
| Use MTTR rather than the reported ensemble baseline | RVOS · Refer-YouTube-VOS public validation | J&F: 54.8 -> 55.32 | Helped |
| Increase the input window from w=1 to w=10 (ablation) | MTTR · A2D-Sentences | mAP: 41.8 -> 46.1 | Helped |
| Increase the input window from w=10 to w=12 (ablation) | MTTR · A2D-Sentences | mAP: 46.1 -> 44.0 | Hurt |
| Replace Video Swin-T with DeepLab-ResNet101 and use target frames only (ablation) | MTTR · A2D-Sentences | mAP: 41.8 -> 41.2 | Hurt |
| Replace RoBERTa-base with another Transformer text encoder (ablation) | MTTR · A2D-Sentences | mAP: 44.0; 44.3; 43.8 | No clear effect |
| Replace RoBERTa-base with GloVe embeddings (ablation) | MTTR · A2D-Sentences | mAP: 44.0 -> 43.4 | Hurt |
| Replace RoBERTa-base with fastText embeddings (ablation) | MTTR · A2D-Sentences | mAP: 44.0 -> 43.1 | Hurt |
| Use 50 object queries rather than 10, 100, 200, or 300 (ablation) | MTTR · A2D-Sentences | mAP: 42.5; 44.0; 41.5; 42.8; 42.7 | Mixed |
| Remove supervision of unreferred instances (ablation) | MTTR · A2D-Sentences | mAP | Mixed |
| Select an instance sequence using temporal segment voting rather than frame-by-frame selection (ablation) | MTTR · Refer-YouTube-VOS | segmentation correctness | Helped |
Closest papers by meaning
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.93)
- SAMWISE: Infusing Wisdom in SAM2 for Text-Driven Video Segmentation (CVPR 2025 · similarity 0.92)
- Semantic and Sequential Alignment for Referring Video Object Segmentation (CVPR 2025 · similarity 0.91)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.91)
- DTOS: Dynamic Time Object Sensing with Large Multimodal Model (CVPR 2025 · similarity 0.89)
- ViCaS: A Dataset for Combining Holistic and Pixel-level Video Understanding using Captions with Grounded 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.