Research map / Vision-language and multimodal / Human and embodied modeling
Towards Real-Time Panoptic Narrative Grounding by an End-to-End Grounding Network
Haowei Wang, Jiayi Ji, Yiyi Zhou, Yongjian Wu, Xiaoshuai Sun
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 LPA and SAL to the one-stage baseline | EPNG · PNG test | Average Recall (All): 40.3 -> 49.7 | Improved |
| Use EPNG instead of the two-stage PNG model | PNG · PNG test | Inference time: 107ms -> 11ms | Improved |
| Train EPNG with the same data as PNG | EPNG · PNG test | Average Recall (All): 49.7 -> 58.0 | Improved |
| Add LPA rather than using no position embedding (ablation) | EPNG · PNG test | Average Recall (All): 42.8 -> 49.7 | Improved |
| Use LPA instead of PE (ablation) | EPNG · PNG test | Average Recall (All): 46.7 -> 49.7 | Improved |
| Use LPA instead of SPE (ablation) | EPNG · PNG test | Average Recall (All): 46.1 -> 49.7 | Improved |
| Use LPA instead of DPE (ablation) | EPNG · PNG test | Average Recall (All): 45.9 -> 49.7 | Improved |
| Use LPA instead of RPE (ablation) | EPNG · PNG test | Average Recall (All): 45.3 -> 49.7 | Improved |
| Add SAL to BCE + Dice loss (ablation) | EPNG · PNG test | Average Recall (All): 43.7 -> 49.7 | Improved |
| Transfer EPNG zero-shot rather than use a random model | EPNG · RefCOCO testA | mIoU: 9.2 -> 28.0 | Improved |
| Transfer EPNG zero-shot rather than use a random model | EPNG · RefCOCO testB | mIoU: 9.8 -> 17.7 | Improved |
| Transfer EPNG zero-shot rather than use a random model | EPNG · RefCOCO+ testA | mIoU: 9.3 -> 27.7 | Improved |
Closest papers by meaning
- Improving Panoptic Narrative Grounding by Harnessing Semantic Relationships and Visual Confirmation (AAAI 2024 · similarity 0.94)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.89)
- Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Grounding (CVPR 2025 · similarity 0.89)
- Look Around Before Locating: Considering Content and Structure Information for Visual Grounding (AAAI 2025 · similarity 0.88)
- When Visual Grounding Meets Gigapixel-level Large-scale Scenes: Benchmark and Approach (CVPR 2024 · similarity 0.87)
- GeM-VG: Towards Generalized Multi-image Visual Grounding with Multimodal Large Language Models (AAAI 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.
