Research map / Video / Video-language understanding
RGNet: A Unified Clip Retrieval and Grounding Network for Long Videos
Tanveer Hannan, Md Mohaiminul Islam, Thomas Seidl, Gedas Bertasius
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 |
|---|---|---|---|
| unify clip retrieval and grounding in one RG-Encoder instead of a disjoint pipeline (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 14.15 -> 18.28 | Improved |
| add learned sparsifier for frame selection (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 16.12 -> 17.80 | Improved |
| add retrieval token for clip features (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 17.80 -> 18.28 | Improved |
| add attention loss (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 16.21 -> 18.28 | Improved |
| add contrastive loss with long-video negative clip sampling (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 17.41 -> 18.28 | Improved |
| learned retrieval instead of heuristic dot-product retrieval (ablation) | RGNet · Ego4D-NLQ val | Clip retrieval R1: 12.41 -> 25.01 | Improved |
| unified architecture improves both retrieval and oracle grounding | RGNet · MAD | Oracle grounding R1@0.3: 29.49 -> 33.42 | Improved |
| pre-train with NaQ annotations (ablation) | RGNet · Ego4D-NLQ val | R1@0.3: 18.28 -> 20.63 | Improved |
| lower frame rate as heuristic sparsification (ablation) | RGNet · Ego4D-NLQ val | grounding accuracy | Hurt |
| increase clip length beyond 180s (MAD) or 48s (Ego4D) (ablation) | RGNet · MAD / Ego4D-NLQ | grounding accuracy | Hurt |
| increase decoder queries beyond 5 (ablation) | RGNet · Ego4D-NLQ val | LVTG performance | Hurt |
| RGNet vs SOONet on MAD R5@0.3 | RGNet · MAD | R5@0.3: 19.64 -> 18.72 | Hurt |
Later papers that compared with this method
5 later accepted papers report RGNet as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- HieraMamba: Video Temporal Grounding via Hierarchical Anchor-Mamba Pooling (CVPR 2026 · better on 2 of 2 shared results)
- ReVisionLLM: Recursive Vision-Language Model for Temporal Grounding in Hour-Long Videos (CVPR 2025 · better on 3 of 3 shared results)
- DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos (CVPR 2025 · better on 2 of 2 shared results)
- Object-Shot Enhanced Grounding Network for Egocentric Video (CVPR 2025 · better on 2 of 2 shared results)
- Universal Video Temporal Grounding with Generative Multi-modal Large Language Models (NeurIPS 2025 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos (CVPR 2025 · similarity 0.89)
- MomentSeeker: A Task-Oriented Benchmark For Long-Video Moment Retrieval (NeurIPS 2025 · similarity 0.88)
- Fewer Steps, Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language (AAAI 2024 · similarity 0.87)
- Generative Video Diffusion for Unseen Novel Semantic Video Moment Retrieval (AAAI 2025 · similarity 0.86)
- When One Moment Isn't Enough: Multi-Moment Retrieval with Cross-Moment Interactions (NeurIPS 2025 · similarity 0.86)
- Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection (CVPR 2024 · 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.
