Research map / Detection and segmentation / CNN features
Language As Queries for Referring Video Object Segmentation
Jiannan Wu, Yi Jiang, Peize Sun, Zehuan Yuan, Ping Luo
Check 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 |
|---|---|---|---|
| Use ReferFormer rather than URVOS with ResNet-50 | Referring video object segmentation · Ref-Youtube-VOS | J&F: 47.2 -> 55.6 | Improved |
| Use ReferFormer rather than URVOS with ResNet-50 | Referring video object segmentation · Ref-DAVIS17 | J&F: 51.5 -> 58.5 | Improved |
| Use ReferFormer with Video-Swin-B rather than MTTR with Video-Swin-T | Referring video object segmentation · A2D-Sentences | mAP: 46.1 -> 55.0 | Improved |
| Use ReferFormer with Video-Swin-B rather than MTTR with Video-Swin-T | Referring video object segmentation · JHMDB-Sentences | mAP: 39.2 -> 43.7 | Improved |
| Add visual-language fusion (ablation) | ReferFormer · Ref-Youtube-VOS (Video-Swin-T, trained from scratch) | J: 53.0 -> 54.8 | Improved |
| Add relative coordinates to mask features (ablation) | ReferFormer · Ref-Youtube-VOS (Video-Swin-T, trained from scratch) | J: 53.7 -> 54.8 | Improved |
| Use 5 rather than 8 conditional queries (ablation) | ReferFormer · Ref-Youtube-VOS (Video-Swin-T, trained from scratch) | J&F: 55.3 -> 56.0 | Improved |
| Train with 5 rather than 1 frame per clip (ablation) | ReferFormer · Ref-Youtube-VOS (Video-Swin-T, trained from scratch) | J&F: 50.0 -> 56.0 | Improved |
| Include box and mask costs in label assignment (ablation) | ReferFormer · Ref-Youtube-VOS (Video-Swin-T, trained from scratch) | J&F: 55.2 -> 56.0; 54.5 -> 56.0 | Improved |
| Use class-agnostic rather than class-discriminative training (ablation) | ReferFormer · Ref-Youtube-VOS | J&F: 53.9 -> 56.0 | Improved |
| Add CFBI post-processing to the model (ablation) | ReferFormer with ResNet-50 · Ref-Youtube-VOS | J&F: 55.6 -> 59.4 | Improved |
| Add CFBI post-processing to the model (ablation) | ReferFormer with Swin-L · Ref-Youtube-VOS | J&F: 62.4 -> 63.3 | Improved |
Closest papers by meaning
- Temporal Collection and Distribution for Referring Video Object Segmentation (ICCV 2023 · similarity 0.92)
- End-to-End Referring Video Object Segmentation With Multimodal Transformers (CVPR 2022 · similarity 0.92)
- LoSh: Long-Short Text Joint Prediction Network for Referring Video Object Segmentation (CVPR 2024 · similarity 0.90)
- Referred by Multi-Modality: A Unified Temporal Transformer for Video Object Segmentation (AAAI 2024 · similarity 0.89)
- Point-VOS: Pointing Up 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.
