Research map / 3D vision / 3D generation
ISNN: Impact Sound Neural Network for Audio-Visual Object Classification
Auston Sterling, Justin Wilson, Sam Lowe, Ming C. Lin
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 synthetic sounds to real sounds for training | ISNN-A · RSAudio real | accuracy (%): up to 11 % | Improved |
| train on merged real and synthetic sounds instead of real only (ablation) | ISNN-A · RSAudio | accuracy (%): 92.37% -> 97.07% | Improved |
| add acoustic augmentations (room acoustics, noise, time variance) to synthetic sounds (ablation) | ISNN-A · RSAudio real, trained on synthetic | accuracy (%) | No clear effect |
| add voxel (VoxNet) input to audio (ablation) | ISNN-A · MN10o | accuracy (%): 43.35% -> 91.80% | Improved |
| add voxel input to audio on MN10os | ISNN-AV · MN10os | accuracy (%): 84.50% | No clear effect |
| replace concatenation with bilinear pooling (MFB) (ablation) | ISNN-AV · MN10osm | accuracy (%): 88.50% -> 90.12% | Improved |
| replace addition with bilinear pooling (MFB) (ablation) | ISNN-AV · MN10osm | accuracy (%): 86.00% -> 90.12% | Improved |
| replace multiplicative fusion with MFB (ablation) | ISNN-AV · MN40osm | accuracy (%): 93.24% -> 92.51% | Hurt |
| replace SoundNet8 with ISNN-A as audio branch in MFB fusion (ablation) | ISNN-AV · MN10osm | accuracy (%): 42.00% -> 90.12% | Improved |
| assign material and scale to classes (osm vs o) (ablation) | ISNN-A · MN10 | accuracy (%): 43.35% -> 71.50% | Improved |
| assign material and scale to classes (ablation) | ISNN-AV (MFB) · MN10 | accuracy (%): 91.80% -> 90.12% | Hurt |
| assign scale and material to ModelNet40 classes (ablation) | ISNN-A · MN40 | accuracy (%): 32.51% -> 65.07% | Improved |
Closest papers by meaning
- Objects as Audio-Visual Modal Sound Fields (ECCV 2026 · similarity 0.86)
- RealImpact: A Dataset of Impact Sound Fields for Real Objects (CVPR 2023 · similarity 0.85)
- NeRAF: 3D Scene Infused Neural Radiance and Acoustic Fields (ICLR 2025 · similarity 0.82)
- Audio Scanning Network: Bridging Time and Frequency Domains for Audio Classification (AAAI 2024 · similarity 0.82)
- Improving Audio-Visual Segmentation with Bidirectional Generation (AAAI 2024 · similarity 0.82)
- Dynamic Derivation and Elimination: Audio Visual Segmentation with Enhanced Audio Semantics (CVPR 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.
