Research map / 3D vision / Face and shape
INRAS: Implicit Neural Representation for Audio Scenes
Kun Su, Mingfei Chen, Eli Shlizerman
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 |
|---|---|---|---|
| INRAS vs NAF | INRAS · SoundSpaces single scene | C50 error (dB): 1.06 -> 0.6 | Improved |
| INRAS vs NAF | INRAS · SoundSpaces single scene | EDT error (s): 0.031 -> 0.019 | Improved |
| INRAS vs NAF, model size | INRAS · SoundSpaces single scene | inference time (ms): 37.86 -> 9.47 | Improved |
| train one INRAS on three scenes (multi-condition) | INRAS · SoundSpaces multi-scene | C50 error (dB): 1.05 -> 0.68 | Improved |
| multi-condition training vs NAF on T60 | INRAS · SoundSpaces multi-scene | T60 error (%): 3.01 -> 4.09 | Hurt |
| Simple INRAS with MSE loss only (remove architecture) (ablation) | INRAS · representative scene | T60 error (%): 2.07 -> 49.6 | Hurt |
| add multi-resolution STFT loss to Simple INRAS (ablation) | Simple INRAS · representative scene | T60 error (%): 49.6 -> 6.40 | Improved |
| add multi-resolution STFT loss to Simple INRAS (ablation) | Simple INRAS · representative scene | C50 error (dB): 1.47 -> 2.20 | Hurt |
| remove relative distance input (ablation) | INRAS · representative scene | C50 error (dB): 0.44 -> 1.12 | Hurt |
| remove bounce module (ablation) | INRAS · representative scene | C50 error (dB): 0.44 -> 0.63 | Hurt |
| incomplete bounce point coverage (less) (ablation) | INRAS · representative scene | T60 error (%): 2.07 -> 2.17 | Hurt |
| incomplete bounce point coverage (more) (ablation) | INRAS · representative scene | T60 error (%): 2.07 -> 2.31 | Hurt |
Later papers that compared with this method
5 later accepted papers report INRAS as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- INFER: Learning Implicit Neural Frequency Response Fields for Confined Acoustic Environments (ICML 2026 · better on 8 of 8 shared results)
- NeRAF: 3D Scene Infused Neural Radiance and Acoustic Fields (ICLR 2025 · better on 3 of 3 shared results)
- Resounding Acoustic Fields with Reciprocity (NeurIPS 2025 · better on 1 of 1 shared result)
- Differentiable Room Acoustic Rendering with Multi-View Vision Priors (ICCV 2025 · better on 1 of 1 shared result)
- AV-NeRF: Learning Neural Fields for Real-World Audio-Visual Scene Synthesis (NeurIPS 2023 · better on 5 of 5 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- NeRAF: 3D Scene Infused Neural Radiance and Acoustic Fields (ICLR 2025 · similarity 0.88)
- Representing Sounds as Neural Amplitude Fields: A Benchmark of Coordinate-MLPs and a Fourier Kolmogorov-Arnold Framework (AAAI 2025 · similarity 0.86)
- AV-GS: Learning Material and Geometry Aware Priors for Novel View Acoustic Synthesis (NeurIPS 2024 · similarity 0.86)
- Hearing Anything Anywhere (CVPR 2024 · similarity 0.86)
- Resounding Acoustic Fields with Reciprocity (NeurIPS 2025 · similarity 0.86)
- OWL : Geometry-Aware Spatial Reasoning for Audio Large Language Models (ICLR 2026 · 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.
