Research map / Video / Video-language understanding
EgoCVR: An Egocentric Benchmark for Fine-Grained Composed Video Retrieval
Thomas Hummel, Shyamgopal Karthik, Mariana-Iuliana Georgescu, Zeynep Akata
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 visual-candidate filtering before TF-CVR text re-ranking (ablation) | TF-CVR · EgoCVR global | Recall@1: 4.4 -> 14.1 | Improved |
| use LanguageBind instead of EgoVLPv2 for the visual filtering stage (ablation) | TFR-CVR · EgoCVR global | Recall@1: 12.2 -> 14.1 | Improved |
| process the whole video instead of a single middle frame (ablation) | TFR-CVR · EgoCVR global | Recall@5: 27.4 -> 39.5 | Improved |
| use full video instead of a single frame (ablation) | LanguageBind · EgoCVR global | Recall@1: 6.4 -> 6.1 | No clear effect |
| replace instruction with LLM-reformulated predicted caption as text query (ablation) | TFR-CVR · EgoCVR global | Recall@1: 12.8 -> 14.1 | Improved |
| use ground-truth captions instead of predicted captions (ablation) | TFR-CVR · EgoCVR local | Recall@1: 44.2 -> 51.7 | Improved |
| use too many re-ranking candidates (n_c > 30) | TFR-CVR · EgoCVR global | Recall | Hurt |
| apply LanguageBind re-ranking to BLIP_CoVR (ablation) | BLIP_CoVR · EgoCVR global | Recall@1: 5.4 -> 10.4 | Improved |
| apply LanguageBind re-ranking to CLIP (ablation) | CLIP · EgoCVR global | Recall@10: 55.6 -> 54.4 | No clear effect |
| finetune BLIP on WebVid-CoVR (BLIP_CoVR) versus zero-shot methods | BLIP_CoVR · EgoCVR global | Recall@1: 8.7 -> 5.4 (BLIP Avg vs BLIP_CoVR) | Hurt |
| add visual filtering to TF-CVR (ablation) | TF-CVR · WebVid-CoVR-Test | Recall@1: 48.4 -> 51.7 | Improved |
Closest papers by meaning
- From Play to Replay: Composed Video Retrieval for Temporally Fine-Grained Videos (NeurIPS 2025 · similarity 0.90)
- MUVR: A Multi-Modal Untrimmed Video Retrieval Benchmark with Multi-Level Visual Correspondence (NeurIPS 2025 · similarity 0.89)
- Composed Video Retrieval via Enriched Context and Discriminative Embeddings (CVPR 2024 · similarity 0.88)
- CoVR: Learning Composed Video Retrieval from Web Video Captions (AAAI 2024 · similarity 0.87)
- Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension (NeurIPS 2025 · similarity 0.87)
- AdaVideoRAG: Omni-Contextual Adaptive Retrieval-Augmented Efficient Long Video Understanding (NeurIPS 2025 · 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.
