Research map / Vision-language and multimodal / Cross-modal learning
Sep-Stereo: Visually Guided Stereophonic Audio Generation by Associating Source Separation
Hang Zhou, Xudong Xu, Dahua Lin, Xiaogang Wang, Ziwei Liu
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 Associative-Conv visual-audio fusion to Mono2Binaural (ablation) | Mono2Binaural · FAIR-Play | STFT distance: 0.959 -> 0.893 | Improved |
| add Associative-Conv visual-audio fusion to Mono2Binaural (ablation) | Mono2Binaural · YT-Music | STFT distance: 1.346 -> 1.147 | Improved |
| expand Associative-Conv to multi-scale APNet (ablation) | Sep-Stereo · YT-Music | STFT distance: 1.147 -> 1.070 | Improved |
| expand Associative-Conv to multi-scale APNet (ablation) | Sep-Stereo · FAIR-Play | STFT distance: 0.893 -> 0.889 | Improved |
| add separative learning with mono MUSIC data to APNet (ablation) | Sep-Stereo · FAIR-Play | STFT distance: 0.889 -> 0.879 | Improved |
| add separative learning with mono MUSIC data to APNet (ablation) | Sep-Stereo · YT-Music | STFT distance: 1.070 -> 1.051 | Improved |
| add separative learning to Mono2Binaural via feature rearrangement (ablation) | Mono2Binaural · FAIR-Play | STFT distance: 0.959 -> 0.930 | Improved |
| add output-difference loss L_D (ablation) | Sep-Stereo · FAIR-Play | STFT distance: 0.976 -> 0.879 | Improved |
| use APNet left/right branch outputs instead of direct U-Net difference output (ablation) | Sep-Stereo · FAIR-Play | STFT distance: 0.933 -> 0.879 | Improved |
| use horizontal placement instead of vertical extremes in feature rearrangement (ablation) | Sep-Stereo · FAIR-Play | STFT distance: 0.919 -> 0.879 | Improved |
| add APNet-based separation with shared backbone versus Associative-Conv (ablation) | Sep-Stereo · MUSIC source separation | SDR: 5.79 -> 8.07 | Improved |
| Sep-Stereo versus PixelPlayer for separation | Sep-Stereo · MUSIC source separation | SDR/SIR/SAR: SDR 7.67 -> 8.07, SIR 14.81 -> 10.14, SAR 11.24 -> 15.51 | Mixed |
Closest papers by meaning
- Visually Informed Binaural Audio Generation without Binaural Audios (CVPR 2021 · similarity 0.90)
- 2.5D Visual Sound (CVPR 2019 · similarity 0.90)
- Localize to Binauralize: Audio Spatialization From Visual Sound Source Localization (ICCV 2021 · similarity 0.88)
- Both Ears Wide Open: Towards Language-Driven Spatial Audio Generation (ICLR 2025 · similarity 0.86)
- Cyclic Learning for Binaural Audio Generation and Localization (CVPR 2024 · similarity 0.86)
- V2A-Mapper: A Lightweight Solution for Vision-to-Audio Generation by Connecting Foundation Models (AAAI 2024 · 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.
