Research map / Video / Video-language understanding
Music Grounding by Short Video
Zijie Xin, Minquan Wang, Jingyu Liu, Quan Chen, Ye Ma, Peng Jiang, Xirong Li
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 |
|---|---|---|---|
| remove cross-attention from fusion (ablation) | QD-DETR · single-music | mIoU: 0.423 -> 0.634 | Improved |
| remove cross-attention from fusion (ablation) | TR-DETR · single-music | mIoU: 0.393 -> 0.630 | Improved |
| remove dual cross-attention (ablation) | UVCOM · single-music | mIoU: 0.652 -> 0.661 | Improved |
| add SA feature enhancement (ablation) | MaDe · music-set | MoR1: 6.4 -> 8.3 | Improved |
| add X-Pool embedding to mean pooling (ablation) | MaDe · music-set | MoR1: 7.1 -> 8.3 | Improved |
| add mean-pooled embedding to X-Pool (ablation) | MaDe · music-set | MoR1: 6.3 -> 8.3 | Improved |
| use joint loss instead of single loss (ablation) | MaDe · music-set | MoR1: 7.5 -> 8.3 | Improved |
| combine by similarity rather than feature addition (ablation) | MaDe · music-set | MoR1: 6.9 -> 8.3 | Improved |
| use video embedding as initial content token vs zero (ablation) | MaDe · single-music | mIoU: 0.709 -> 0.722 | Improved |
| use 1 query token instead of 10 (ablation) | MaDe · single-music | mIoU: 0.716 -> 0.722 | Improved |
| predict width p_w in addition to center (ablation) | MaDe · single-music | mIoU: 0.706 -> 0.722 | Improved |
| use SA fusion instead of CA (ablation) | MaDe · single-music | mIoU: 0.697 -> 0.722 | Improved |
Closest papers by meaning
- GVMGen: A General Video-to-Music Generation Model with Hierarchical Attentions (AAAI 2025 · similarity 0.84)
- V2Meow: Meowing to the Visual Beat via Video-to-Music Generation (AAAI 2024 · similarity 0.83)
- VidMuse: A Simple Video-to-Music Generation Framework with Long-Short-Term Modeling (CVPR 2025 · similarity 0.82)
- Supervised Contrastive Learning from Weakly-Labeled Audio Segments for Musical Version Matching (ICML 2025 · similarity 0.82)
- Fewer Steps, Better Performance: Efficient Cross-Modal Clip Trimming for Video Moment Retrieval Using Language (AAAI 2024 · similarity 0.82)
- Empower Words: DualGround for Structured Phrase and Sentence-Level Temporal Grounding (NeurIPS 2025 · similarity 0.81)
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.
