Research map / Video / Video-language understanding
Query-Focused Extractive Video Summarization
Aidean Sharghi, Boqing Gong, Mubarak Shah
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 |
|---|---|---|---|
| use SH-DPP (query layer plus sequential summary layer) instead of vanilla DPP | query-focused summarizer · TV episodes, patient user, bi-concept queries | F-measure (%): 29.62 -> 37.02 | Improved |
| use SH-DPP instead of seqDPP | query-focused summarizer · UTE, patient user, bi-concept queries | F-measure (%): 18.85 -> 21.27 | Mixed |
| use SH-DPP instead of seqDPP | query-focused summarizer · UTE, impatient user, bi-concept queries | hitting recall (%): 12.68 -> 22.91 | Improved |
| use SH-DPP instead of uniform sampling | query-focused summarizer · UTE, impatient user, bi-concept queries | F-measure (%): 25.44 -> 25.56 | No clear effect |
| use SH-DPP instead of SubMod | query-focused summarizer · UTE, impatient user, bi-concept queries | F-measure (%): 27.10 -> 25.56 | Hurt |
| use SH-DPP trained on bi-concept queries for novel three-concept queries (vs DPP) | SH-DPP · UTE, patient user, three-concept queries | F-measure (%): 20.7 -> 24.54 | Improved |
| use SH-DPP instead of seqDPP on three-concept queries | SH-DPP · UTE, impatient user, three-concept queries | F-measure (%): 19.4 -> 29.59 | Improved |
| add dependency between adjacent subsets (seqDPP) vs DPP on long egocentric UTE videos (ablation) | DPP · UTE, patient user, bi-concept queries | F-measure (%): 15.7 -> 18.85 | Improved |
| add dependency between adjacent subsets (seqDPP) vs DPP on TV episodes (ablation) | DPP · TV episodes, patient user, bi-concept queries | F-measure (%): 29.62 -> 27.96 | No clear effect |
| vary ground set size from 6 to 10 shots (ablation) | SH-DPP · TV episodes | F-measure | No clear effect |
| use very small (4 shots) or large (16 shots) ground set (ablation) | SH-DPP · TV episodes | F-measure | Hurt |
| vary lower-dimension of W and V (ablation) | SH-DPP · UTE and TV | F-measure | No clear effect |
Closest papers by meaning
- Query-Focused Video Summarization: Dataset, Evaluation, and a Memory Network Based Approach (CVPR 2017 · similarity 0.92)
- CLIP-It! Language-Guided Video Summarization (NeurIPS 2021 · similarity 0.89)
- IntentVizor: Towards Generic Query Guided Interactive Video Summarization (CVPR 2022 · similarity 0.88)
- Video Summarization with Large Language Models (CVPR 2025 · similarity 0.88)
- Video Summarization Using Denoising Diffusion Probabilistic Model (AAAI 2025 · similarity 0.86)
- Scaling Up Video Summarization Pretraining with Large Language Models (CVPR 2024 · similarity 0.85)
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.
