Research map / 3D vision / Face and shape
Shape and Material from Sound
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 self-supervised network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | specific modulus accuracy: 56% -> 62% | Helped |
| Add self-supervised network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | shape accuracy: 54% -> 52% | No clear effect |
| Add self-supervised network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | Rayleigh beta MSE: 0.173 -> 0.117 | Helped |
| Add weak-label network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | shape accuracy: 54% -> 62% | Helped |
| Add weak-label network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | specific modulus accuracy: 56% -> 66% | Helped |
| Add weak-label network initialization (ablation) | unsupervised analysis-by-synthesis · Synthetic falling objects, 80 MCMC sweeps | height MSE: 0.003 -> 0.005 | Hurt |
| Add self-supervised initialization (ablation) | unsupervised analysis-by-synthesis · Likelihood over MCMC sweeps | convergence speed | Helped |
| Use full ground-truth supervision instead of unsupervised inference | recognition model · Synthetic falling objects, 80 MCMC sweeps | shape accuracy: 54% -> 98% | Helped |
| Apply amplitude cutoff in audio synthesis (ablation) | audio engine · 200-collision synthesis | time (s): 30.4 -> 24.5 | Helped |
| Select principal modes (90% energy) (ablation) | audio engine · 200-collision synthesis | time (s): 30.4 -> 12.7 | Helped |
| Multi-threading collisions (ablation) | audio engine · 200-collision synthesis | time (s): 30.4 -> 1.5 | Helped |
| Combine all speedups (ablation) | audio engine · 200-collision synthesis | time (s): 30.4 -> 0.8 | Helped |
Closest papers by meaning
- Visually Indicated Sounds (CVPR 2016 · similarity 0.87)
- RealImpact: A Dataset of Impact Sound Fields for Real Objects (CVPR 2023 · similarity 0.86)
- Hearing Hands: Generating Sounds from Physical Interactions in 3D Scenes (CVPR 2025 · similarity 0.85)
- Generative Perception of Shape and Material from Differential Motion (NeurIPS 2025 · similarity 0.84)
- Physical Property Understanding from Language-Embedded Feature Fields (CVPR 2024 · similarity 0.82)
- Teaching Physical Awareness to LLMs through Sounds (ICML 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.