Research map / Detection and segmentation / CNN features
Learning Trailer Moments in Full-Length Movies with Co-Contrastive Attention
Lezi Wang, Dong Liu, Rohit Puri, Dimitris N. Metaxas
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 Co-Attention soft-label ranking instead of fully supervised hard labels | trailer moment ranker · TMDD (Average) | Rank@Global AP: 0.151 -> 0.208 | Improved |
| use Co-Attention instead of offline pseudo labels | trailer moment ranker · TMDD (Average) | Rank@Global AP: 0.193 -> 0.208 | Improved |
| add Contrastive Attention feature augmentation | supervised ranker · TMDD (Average) | Rank@Global AP: 0.151 -> 0.162 | Improved |
| add Contrastive Attention feature augmentation | pseudo-label ranker · TMDD (Average) | Rank@Global AP: 0.193 -> 0.232 | Improved |
| add Contrastive Attention feature augmentation and loss (ablation) | Co-Attention ranker · TMDD (Average) | Rank@Global AP: 0.208 -> 0.242 | Improved |
| add Contrastive Attention features (ablation) | Co-Attention ranker · TMDD (Average) | Rank@10 AP: 0.639 -> 0.669 | Improved |
| add feature augmentation (Eq. 5) without contrastive loss (ablation) | CoA · TMDD (Action) | Rank@Global AP: 0.228 -> 0.255 | Improved |
| add feature augmentation (Eq. 5) without contrastive loss (ablation) | CoA · TMDD (Sci-Fi) | Rank@Global AP: 0.208 -> 0.236 | Improved |
| add contrastive loss LC on top of feature augmentation (ablation) | CoA+FeaAug · TMDD (Action) | Rank@Global AP: 0.255 -> 0.271 | Improved |
| add contrastive loss LC on top of feature augmentation (ablation) | CoA+FeaAug · TMDD (Drama) | Rank@Global AP: 0.230 -> 0.246 | Improved |
| add contrastive loss LC on top of feature augmentation (ablation) | CoA+FeaAug · TMDD (Sci-Fi) | Rank@Global AP: 0.236 -> 0.242 | Improved |
| vary scaling factor lambda (ablation) | CCANet · TMDD | Rank@10 AP | No clear effect |
Closest papers by meaning
- Towards Automated Movie Trailer Generation (CVPR 2024 · similarity 0.84)
- Multi-Scale Contrastive Learning for Video Temporal Grounding (AAAI 2025 · similarity 0.83)
- Watch Video, Catch Keyword: Context-aware Keyword Attention for Moment Retrieval and Highlight Detection (AAAI 2025 · similarity 0.83)
- Generative Video Diffusion for Unseen Novel Semantic Video Moment Retrieval (AAAI 2025 · similarity 0.82)
- MomentSeeker: A Task-Oriented Benchmark For Long-Video Moment Retrieval (NeurIPS 2025 · similarity 0.82)
- TeaserGen: Generating Teasers for Long Documentaries (ICLR 2025 · similarity 0.82)
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.
