Research map / Video / Video-language understanding
G2L: Semantically Aligned and Uniform Video Grounding via Geodesic and Game Theory
Hongxiang Li, Meng Cao, Xuxin Cheng, Yaowei Li, Zhihong Zhu, Yuexian Zou
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 geodesic-guided contrastive learning (GCL) (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.7: 28.62 -> 33.35 | Improved |
| add semantic alignment term in GCL (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.7: 30.73 -> 33.35 | Improved |
| add semantic uniformity term in GCL (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.7: 31.14 -> 33.35 | Improved |
| add semantic Shapley interaction (SSI) (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.7: 30.01 -> 33.35 | Improved |
| add GCL and SSI over the baseline (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.5: 45.74 -> 51.68 | Improved |
| use geodesic distance instead of Euclidean distance (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.5: 47.83 -> 51.68 | Improved |
| use geodesic distance instead of timestamp distance (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.5: 49.85 -> 51.68 | Improved |
| use geodesic distance instead of cosine distance (ablation) | G2L · ActivityNet-Captions | R@1 IoU=0.7: 30.09 -> 33.35 | Improved |
| add GCL and SSI training objectives | Base model · ActivityNet-Captions | training iteration time: 0.39s -> 0.84s | Hurt |
| add GCL and SSI training objectives | Base model · ActivityNet-Captions | inference time: 40s -> 43s | Hurt |
| G2L versus best prior method on TACoS | G2L · TACoS | R@5 IoU=0.5: 50.00 -> 49.86 | Mixed |
Closest papers by meaning
- SnAG: Scalable and Accurate Video Grounding (CVPR 2024 · similarity 0.88)
- VideoGrounding-DINO: Towards Open-Vocabulary Spatio-Temporal Video Grounding (CVPR 2024 · similarity 0.87)
- STPro: Spatial and Temporal Progressive Learning for Weakly Supervised Spatio-Temporal Grounding (CVPR 2025 · similarity 0.87)
- Multi-Scale Contrastive Learning for Video Temporal Grounding (AAAI 2025 · similarity 0.87)
- VideoGEM: Training-free Action Grounding in Videos (CVPR 2025 · similarity 0.87)
- Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Video Temporal Grounding (NeurIPS 2025 · similarity 0.87)
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.
