Research map / Video / Video-language understanding
Decoupled Seg Tokens Make Stronger Reasoning Video Segmenter and Grounder
Jisheng Dang, Xudong Wu, Haowen Yan, Huicheng Zheng, Wei‐Shi Zheng, Jianhuang Lai, Bin Hu, Tat‐Seng Chua
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 text understanding pre-training of SAM-2 (ablation) | DeSa2VA-4B · RefCOCO | cIoU: 79.2 -> 82.6 | Improved |
| add text understanding pre-training of SAM-2 (ablation) | DeSa2VA-4B · MeViS | J&F: 45.9 -> 47.2 | Improved |
| add text pre-training to the Sa2VA baseline without decoupling (ablation) | Sa2VA-4B · RefCOCO | cIoU: 78.9 -> 73.4 | Hurt |
| add text pre-training to the Sa2VA baseline without decoupling (ablation) | Sa2VA-1B · RefCOCO | cIoU: 77.4 -> 72.8 | Hurt |
| add text pre-training with decoupling (ablation) | DeSa2VA-1B · MeViS | J&F: 40.9 -> 45.1 | Improved |
| add mask reprompting (dense mask prompt) (ablation) | DeSa2VA-4B · RefCOCO | cIoU: 81.3 -> 82.6 | Improved |
| add mask reprompting (dense mask prompt) (ablation) | Sa2VA-4B · RefCOCO | cIoU: 78.9 -> 80.9 | Improved |
| add mask reprompting (ablation) | DeSa2VA-4B · SEED-Bench | accuracy: 73.2 -> 73.3 | No clear effect |
| add the decoupling module (text/visual linear projections) to Sa2VA | DeSa2VA-4B vs Sa2VA-4B · RefCOCO+ | cIoU: 71.7 -> 77.8 | Improved |
| add decoupling to Sa2VA on video QA and image QA | DeSa2VA-4B · SEED-Bench | accuracy: 73.3 -> 73.3 | No clear effect |
| add image segmentation data to training (ablation) | DeSa2VA-4B · RefCOCO | cIoU: 21.5 -> 82.6 | Improved |
| add video segmentation data to training (ablation) | DeSa2VA-4B · RefCOCOg | cIoU: 76.8 -> 79.2 | Mixed |
Closest papers by meaning
- Beyond Token Representations: Explicit Visual Object Grounding for Video Reasoning Segmentation (NeurIPS 2026 · similarity 0.90)
- SimVG: A Simple Framework for Visual Grounding with Decoupled Multi-modal Fusion (NeurIPS 2024 · similarity 0.87)
- One Token to Seg Them All: Language Instructed Reasoning Segmentation in Videos (NeurIPS 2024 · similarity 0.87)
- Seg-ReSearch: Segmentation with Interleaved Reasoning and External Search (ICML 2026 · similarity 0.87)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.86)
- VGent: Visual Grounding via Modular Design for Disentangling Reasoning and Prediction (CVPR 2026 · similarity 0.86)
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.
