Research map / 3D vision / 3D generation
How Would It Sound? Material-Controlled Multimodal Acoustic Profile Generation for Indoor Scenes
Mahnoor Fatima Saad, Ziad Al-Halah
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 target material mask M_T as input (ablation) | M-CAPA audio-visual · AcWon D_uu | RTE: 109.46 -> 91.44 | Improved |
| add target material mask M_T as input (ablation) | M-CAPA audio-visual · AcWon D_uu | CTE: 9.19 -> 8.44 | Improved |
| add residual B_T alongside weighting mask W_T (ablation) | M-CAPA audio-visual · AcWon D_uu | STFT: 4.93 -> 3.87 | Improved |
| add residual B_T alongside weighting mask W_T (ablation) | M-CAPA audio-visual · AcWon D_uu | RTE: 105.19 -> 91.44 | Improved |
| use ground-truth simulator segmentation instead of inferred segmentation (ablation) | M-CAPA audio-visual · AcWon D_uu | RTE: 97.63 -> 91.44 | Improved |
| provide full target material mask instead of only changed materials (ablation) | M-CAPA audio-visual · AcWon D_uu | RTE: 96.36 -> 91.44 | Improved |
| add L2 and energy decay loss to L1 loss (ablation) | M-CAPA audio-visual · AcWon | RTE: 97.92 -> 90.61 | Improved |
| add energy decay loss to L1+L2 (ablation) | M-CAPA audio-visual · AcWon | RTE: 109.27 -> 90.61 | Improved |
| add L1 loss to L2-only training (ablation) | M-CAPA audio-visual · AcWon | RTE: 241.41 -> 90.61 | Improved |
| use audio-visual input instead of vision-only (ablation) | M-CAPA audio-visual · AcWon D_uu | STFT: 5.76 -> 3.87 | Improved |
| noise added to source RIR (SNR 40 dB to 0 dB) | M-CAPA audio-visual · AcWon D_uu | STFT, RTE | Hurt |
| M-CAPA versus AV-RIR audio-visual baseline | M-CAPA audio-visual · AcWon D_uu | L1: 7.59 -> 5.27 | Improved |
Closest papers by meaning
- Hearing Anywhere in Any Environment (CVPR 2025 · similarity 0.87)
- Deep Neural Room Acoustics Primitive (ICML 2024 · similarity 0.86)
- Hearing Anything Anywhere (CVPR 2024 · similarity 0.86)
- AV-GS: Learning Material and Geometry Aware Priors for Novel View Acoustic Synthesis (NeurIPS 2024 · similarity 0.85)
- Read, Watch and Scream! Sound Generation from Text and Video (AAAI 2025 · similarity 0.85)
- SoundVista: Novel-View Ambient Sound Synthesis via Visual-Acoustic Binding (CVPR 2025 · similarity 0.85)
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.
