Research map / Video / Video segmentation and tracking
Joint Visual and Audio Learning for Video Highlight Detection
Taivanbat Badamdorj, Mrigank Rochan, Yang Wang, Li Cheng
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 disentangled attention instead of classic self-attention (ablation) | Ours (AV) · VTW | mAP: 0.808 -> 0.812 | Improved |
| add disentangled attention instead of classic self-attention (ablation) | Ours (AV) · YouTube Highlights | mAP: 0.705 -> 0.718 | Improved |
| add noise sentinel to bimodal attention (ablation) | Ours (AV) · VTW | mAP: 0.804 -> 0.812 | Improved |
| add noise sentinel to bimodal attention (ablation) | Ours (AV) · YouTube Highlights | mAP: 0.716 -> 0.718 | Improved |
| add audio via bimodal attention to visual-only self-attention model | SA (V) · VTW | mAP: 0.722 -> 0.812 | Improved |
| add audio via bimodal attention to visual-only self-attention model | SA (V) · YouTube Highlights | mAP: 0.705 -> 0.718 | Improved |
| add audio via bimodal attention to visual-only self-attention model | SA (V) · TVSum | top-5 mAP: 0.7476 -> 0.7627 | Improved |
| use bimodal attention with noise sentinel instead of late fusion of self-attended features | SA (AV)late · VTW | mAP: 0.805 -> 0.812 | Improved |
| use bimodal attention with noise sentinel instead of early fusion | SA (AV)early · YouTube Highlights | mAP: 0.703 -> 0.718 | Improved |
| use audio-only features instead of visual-only features | SA · VTW | mAP: 0.722 -> 0.794 | Improved |
| use audio-only features instead of visual-only features | SA · TVSum | top-5 mAP: 0.748 -> 0.687 | Hurt |
| add audio-visual model over visual-only per category | Ours (AV) · TVSum categories | top-5 mAP: VU 0.6469 -> 0.5726; BT 0.9496 -> 0.9741 | Mixed |
Closest papers by meaning
- Temporal Cue Guided Video Highlight Detection With Low-Rank Audio-Visual Fusion (ICCV 2021 · similarity 0.91)
- Adaptive Video Highlight Detection by Learning from User History (ECCV 2020 · similarity 0.90)
- Bridging the Gap: A Unified Video Comprehension Framework for Moment Retrieval and Highlight Detection (CVPR 2024 · similarity 0.87)
- Learning to Highlight Audio by Watching Movies (CVPR 2025 · similarity 0.87)
- TR-DETR: Task-Reciprocal Transformer for Joint Moment Retrieval and Highlight Detection (AAAI 2024 · similarity 0.87)
- Task-Driven Exploration: Decoupling and Inter-Task Feedback for Joint Moment Retrieval and Highlight Detection (CVPR 2024 · 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.
