Research map / Vision-language and multimodal / Human and embodied modeling
Enhancing Video Summarization via Vision-Language Embedding
Bryan A. Plummer, Matthew Brown, Svetlana Lazebnik
Check 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 semantic representativeness and semantic interestingness | Submod-V · UT Egocentric | F-measure: 29.35 -> 34.15 | Improved |
| Add semantic representativeness (ablation) | Submod-V · UT Egocentric | F-measure: 29.35 -> 33.24 | Improved |
| Add semantic interestingness (ablation) | Submod-V · UT Egocentric | F-measure: 29.35 -> 31.58 | Improved |
| Replace both visual-only representativeness and interestingness with semantic versions (ablation) | Submod-V · UT Egocentric | F-measure: 29.35 -> 29.54 | No clear effect |
| Add semantic representativeness and semantic interestingness (ablation) | Submod-V · UT Egocentric | F-measure: 29.35 -> 33.50 | Improved |
| Train the vision-language embedding on Flickr30k rather than UTE (ablation) | Submod-V + Both · UT Egocentric | F-measure: 34.15 -> 33.50 | Hurt |
| Add semantic representativeness (ablation) | Submod-V · TV Episodes | F-measure: 38.18 -> 40.90 | Mixed |
| Add semantic representativeness (ablation) | Submod-V · TV Episodes | Recall: 33.47 -> 37.02 | Improved |
| Replace visual-only representativeness with semantic representativeness (ablation) | Submod-V · TV Episodes | F-measure: 38.18 -> 37.29 | No clear effect |
| Train the embedding on TV Episodes rather than Flickr30k (ablation) | Submod-V + Sem. Rep. · TV Episodes | F-measure: 39.87 -> 40.90 | Improved |
| Add unconstrained freeform text guidance at test time | Full model with UTE embedding · UT Egocentric | F-measure: 34.15 -> 34.90 | Improved |
| Add unconstrained freeform text guidance at test time | Full model with TV Episodes embedding · TV Episodes | F-measure: 40.90 -> 41.18 | Improved |
Closest papers by meaning
- Query-Focused Video Summarization: Dataset, Evaluation, and a Memory Network Based Approach (CVPR 2017 · similarity 0.89)
- Video Summarization with Large Language Models (CVPR 2025 · similarity 0.89)
- Video-XL: Extra-Long Vision Language Model for Hour-Scale Video Understanding (CVPR 2025 · similarity 0.87)
- Scaling Up Video Summarization Pretraining with Large Language Models (CVPR 2024 · similarity 0.87)
- Distilling Vision-Language Models on Millions of Videos (CVPR 2024 · similarity 0.87)
- V2Xum-LLM: Cross-Modal Video Summarization with Temporal Prompt Instruction Tuning (AAAI 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.
