Research map / Vision-language and multimodal / Cross-modal learning
Image2Reverb: Cross-Modal Reverb Impulse Response Synthesis
Nikhil Singh, Jeff Mentch, Jerry Ng, Matthew Beveridge, Iddo Drori
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 depth maps as a fourth input channel (ablation) | Image2Reverb · Test set T60 estimation (50 epochs) | std of T60 error (%): 83.1 -> 78.8 | Improved |
| add depth maps as a fourth input channel (ablation) | Image2Reverb · Test set T60 estimation (50 epochs) | mean T60 error (%): -9.17 -> -6.03 | Improved |
| add differentiable T60 proxy loss term (ablation) | Image2Reverb · Test set T60 estimation (50 epochs) | std of T60 error (%): 85.97 -> 78.8 | Improved |
| add differentiable T60 proxy loss term (ablation) | Image2Reverb · Test set T60 estimation (50 epochs) | mean T60 error (%): -7.1 -> -6.03 | Improved |
| use Places365-pretrained ResNet50 encoder instead of random init (ablation) | Image2Reverb · Test set T60 estimation (50 epochs) | mean T60 error (%): 43.15 -> -6.03 | Improved |
| generate IR with the model instead of nearest-neighbor retrieval | Image2Reverb · Test set T60 estimation (50 epochs) | std of T60 error (%): 491.02 -> 78.8 | Improved |
| set constant depth map to 0 instead of estimated depth (ablation) | Image2Reverb · Generated IR T60 | std of T60 (s): 1.54 -> 0.87 | Improved |
| set constant depth map to 0.5 instead of estimated depth (ablation) | Image2Reverb · Generated IR T60 | mean T60 (s): 2.07 -> 3.62 | Improved |
| compare generated vs measured IRs on outdoor scenes | Image2Reverb · Expert study (quality rating, Outdoor scenes) | equivalence test p-value: .62 | Hurt |
| compare generated vs measured IRs on large scenes | Image2Reverb · Expert study (quality rating, Large scenes) | equivalence test p-value: <.001 | No clear effect |
Closest papers by meaning
- Hearing Anywhere in Any Environment (CVPR 2025 · similarity 0.87)
- Acoustic Volume Rendering for Neural Impulse Response Fields (NeurIPS 2024 · similarity 0.86)
- Real Acoustic Fields: An Audio-Visual Room Acoustics Dataset and Benchmark (CVPR 2024 · similarity 0.86)
- AV-RIR: Audio-Visual Room Impulse Response Estimation (CVPR 2024 · similarity 0.86)
- Deep Neural Room Acoustics Primitive (ICML 2024 · similarity 0.86)
- Resounding Acoustic Fields with Reciprocity (NeurIPS 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.
