Research map / Video / Video segmentation and tracking
Learning Audio-Visual Dynamics Using Scene Graphs for Audio Source Separation
Moitreya Chatterjee, Narendra Ahuja, Anoop Cherian
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 |
|---|---|---|---|
| replace RBF edge weights with multiscale Chamfer thresholding (ablation) | ASMP full · ASIW | SDR: 9.55 -> 9.23 | Hurt |
| train with direction prediction loss only (lambda1=lambda2=lambda3=0) (ablation) | ASMP full · ASIW | SDR: 9.55 -> 6.4 | Hurt |
| train with direction prediction loss only (lambda1=lambda2=lambda3=0) (ablation) | ASMP full · ASIW | 10-class Dir. Acc.: 42.51 -> 32.7 | Hurt |
| remove direction prediction loss (lambda4=0) (ablation) | ASMP full · ASIW | SDR: 9.6 -> 9.0 | Hurt |
| replace pseudo 3D scene graph with 2D equal-weight graph (no direction loss) (ablation) | ASMP w/o direction · ASIW | SDR: 9.0 -> 8.8 | Hurt |
| replace 3D scene graph with 2D graph while keeping direction loss (ablation) | ASMP full · ASIW | SDR: 9.6 -> 9.1 | Hurt |
| add co-separation loss to direction prediction loss (ablation) | direction-only ASMP · ASIW | SDR: 6.4 -> 7.91 | Improved |
| add orthogonality loss to direction prediction loss (ablation) | direction-only ASMP · ASIW | SDR: 6.4 -> 6.42 | No clear effect |
| add consistency loss to direction prediction loss (ablation) | direction-only ASMP · ASIW | SDR: 6.4 -> 6.39 | No clear effect |
| add motion direction prediction to 2.5D graph model | ASMP · ASIW | SDR: 9.02 -> 9.55 | Improved |
| add motion direction prediction to 2.5D graph model | ASMP · AVE | SDR: 6.52261214 -> 7.18 | Improved |
| increase RBF kernel bandwidth percentile above 25 (ablation) | ASMP · ASIW | SDR | Hurt |
Closest papers by meaning
- Visual Scene Graphs for Audio Source Separation (ICCV 2021 · similarity 0.92)
- Learning to Separate Object Sounds by Watching Unlabeled Video (ECCV 2018 · similarity 0.91)
- Seeing Speech and Sound: Distinguishing and Locating Audio Sources in Visual Scenes (CVPR 2025 · similarity 0.89)
- Improving Audio-Visual Segmentation with Bidirectional Generation (AAAI 2024 · similarity 0.88)
- SoundBrush: Sound as a Brush for Visual Scene Editing (AAAI 2025 · similarity 0.88)
- Dynamic Derivation and Elimination: Audio Visual Segmentation with Enhanced Audio Semantics (CVPR 2025 · similarity 0.88)
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.
