Research map / Vision-language and multimodal / Human and embodied modeling
Improving Generalized Visual Grounding With Instance-Aware Joint Learning
Ming Dai, Wenxuan Cheng, Jiangjiang Liu, Lingfeng Yang, Zhenhua Feng, Wankou Yang, Jingdong Wang
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 multi-task joint training of GREC and GRES (ablation) | InstanceVG baseline · gRefCOCO (ablation) | F1 score / gIoU: 65.98 -> 67.13 F1, 65.03 -> 67.33 gIoU | Improved |
| add anchor point decoder (APD) (ablation) | InstanceVG · gRefCOCO (ablation) | F1 score / gIoU: 67.13 -> 69.17 F1, 67.33 -> 69.24 gIoU | Improved |
| add PIPH (point-instance prediction head) (ablation) | InstanceVG · gRefCOCO (ablation) | F1 score / gIoU: 69.17 -> 71.43 F1, 69.24 -> 72.41 gIoU | Improved |
| add dynamic point selector over top-Nq selection (ablation) | IQG · gRefCOCO (ablation) | F1 score / gIoU: 69.09 -> 71.16 F1, 70.51 -> 71.73 gIoU | Improved |
| add text-injected query (ablation) | IQG · gRefCOCO (ablation) | N-acc: 72.87 -> 75.87 | Improved |
| add text filter alone (ablation) | IQG · gRefCOCO (ablation) | F1 score: 69.20 -> 69.09 | No clear effect |
| use Deformable DETR decoder instead of DETR (ablation) | InstanceVG · gRefCOCO (ablation) | F1 score: 67.87 -> 69.14 | Improved |
| use point-prior reference points in multi-scale deformable decoder (ablation) | InstanceVG · gRefCOCO (ablation) | gIoU: 71.49 -> 72.41 | Improved |
| add instance supervision to global supervision (ablation) | InstanceVG · gRefCOCO (ablation) | gIoU: 67.90 -> 72.15 | Improved |
| add negative-sample supervision (ablation) | InstanceVG · gRefCOCO (ablation) | N-acc: 74.45 -> 75.87 | Improved |
| raise point cost weight from 2.0 to 5.0 (ablation) | point-guided matcher · gRefCOCO (ablation) | F1 score: 71.43 -> 69.90 | Hurt |
| increase query count from 10 to 30 at 1x schedule (ablation) | InstanceVG · gRefCOCO (ablation) | F1 score: 71.43 -> 68.55 | Hurt |
Closest papers by meaning
- SimVG: A Simple Framework for Visual Grounding with Decoupled Multi-modal Fusion (NeurIPS 2024 · similarity 0.89)
- Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints (AAAI 2025 · similarity 0.89)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.89)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.88)
- GroundVLP: Harnessing Zero-Shot Visual Grounding from Vision-Language Pre-training and Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.88)
- UGround: Towards Unified Visual Grounding with Unrolled Transformers (ICML 2026 · 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.
