Research map / Video / Video-language understanding
Beyond Caption-Based Queries in Video Moment Retrieval
David Pujol-Perich, Albert Clapés, Dima Damen, Sergio Escalera, Michael Wray
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 |
|---|---|---|---|
| remove decoder self-attention and add query dropout | CG-DETR · HD-EPIC-S2 | mAPm avg: 20.84 -> 23.93 | Improved |
| remove decoder self-attention and add query dropout | LD-DETR · YC2-S | mAPm avg: 29.10 -> 32.49 | Improved |
| remove decoder self-attention and add query dropout | LD-DETR · HD-EPIC-S3 | Rm avg: 6.13 -> 6.03 | No clear effect |
| remove decoder self-attention and add query dropout | CG-DETR and LD-DETR · ANC-S | mAPm avg: 54.48 -> 56.20 (CG-DETR) | Improved |
| remove self-attention only (no query dropout) (ablation) | CG-DETR · HD-EPIC-S2 | mAPm avg: 20.84 -> 21.02 | No clear effect |
| add query dropout only (keep self-attention) (ablation) | CG-DETR · HD-EPIC-S2 | mAPm avg: 20.84 -> 21.43 | No clear effect |
| combine -SA and QD (ablation) | CG-DETR · HD-EPIC-S2 | # active queries: 3.64 -> 6.43 | Improved |
| set query dropout rate to 0.5 instead of 0.25 (ablation) | CG-DETR -SA+QD · HD-EPIC-S2 | mAPm avg: 23.93 -> 3.84 | Hurt |
| replace 1-to-1 matching with 1-to-k matching (ablation) | CG-DETR -SA+QD · HD-EPIC-S2 | mAPm avg: 23.93 -> 10.39 | Hurt |
| add alternative matching strategies (1-to-5, group, hybrid) to base model (ablation) | CG-DETR · HD-EPIC-S2 | mAPm avg: 20.84 -> 16.30 (1-to-5) | Hurt |
| add data augmentation replicating GT moments (ablation) | CG-DETR · HD-EPIC-S2 | Rm avg: 16.04 -> 13.85 | Hurt |
| add ms matcher (ablation) | CG-DETR · HD-EPIC-S2 | mAPm avg: 20.84 -> 20.94 | No clear effect |
Closest papers by meaning
- Fewer Steps, Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language (AAAI 2024 · similarity 0.93)
- VERIFIED: A Video Corpus Moment Retrieval Benchmark for Fine-Grained Video Understanding (NeurIPS 2024 · similarity 0.91)
- Zero-shot Video Moment Retrieval via Off-the-shelf Multimodal Large Language Models (AAAI 2025 · similarity 0.91)
- Generative Video Diffusion for Unseen Novel Semantic Video Moment Retrieval (AAAI 2025 · similarity 0.90)
- Towards Balanced Alignment: Modal-Enhanced Semantic Modeling for Video Moment Retrieval (AAAI 2024 · similarity 0.89)
- When One Moment Isn't Enough: Multi-Moment Retrieval with Cross-Moment Interactions (NeurIPS 2025 · similarity 0.89)
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.
