Research map / Label-efficient and robust learning / Self-supervised and contrastive
Sound Localization by Self-Supervised Time Delay Estimation
Ziyang Chen, David F. Fouhey, Andrew Owens
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 augmentation (noise, reverb, mixture) (ablation) | StereoCRW on FMA · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.434 -> 0.133 | Improved |
| add augmentation (ablation) | MonoCLR on FMA · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.430 -> 0.187 | Improved |
| add augmentation (ablation) | ZeroNCE on FMA · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.227 -> 0.174 | Improved |
| add augmentation (ablation) | StereoCRW on FAIR · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.241 -> 0.174 | Improved |
| add augmentation (ablation) | StereoCRW on FMA · In-the-wild left/right (885 clips) | accuracy (%): 82.2 -> 87.2 | Improved |
| add augmentation (ablation) | MonoCLR on FMA · In-the-wild left/right (885 clips) | accuracy (%): 83.8 -> 87.4 | Improved |
| add augmentation (ablation) | ZeroNCE on FMA · In-the-wild left/right (885 clips) | accuracy (%): 85.4 -> 85.6 | No clear effect |
| add augmentation (ablation) | Salvati et al. on Vox-Sim · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.126 -> 0.169 | Hurt |
| use cycle-consistent random walk (StereoCRW) instead of slow-feature loss (ZeroNCE) | stereo self-supervised model trained on FMA · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.174 -> 0.133 | Improved |
| train on large FMA (101K clips) instead of FAIR-Play | StereoCRW with augmentation · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.174 -> 0.133 | Improved |
| train on stereo sounds (ZeroNCE) instead of mono (MonoCLR) without augmentation | self-supervised model on FMA · TDE-Simulation (SNR 10, RT60 0.5s) | MAE (ms): 0.430 -> 0.227 | Improved |
| add visual input (AV model) to StereoCRW (ablation) | StereoCRW · Visually-guided TDE (0.96s) | RMSE (ms): 0.493 -> 0.304 | Improved |
Closest papers by meaning
- Mix and Localize: Localizing Sound Sources in Mixtures (CVPR 2022 · similarity 0.90)
- Sound Localization from Motion: Jointly Learning Sound Direction and Camera Rotation (ICCV 2023 · similarity 0.89)
- Self-Supervised Moving Vehicle Tracking With Stereo Sound (ICCV 2019 · similarity 0.88)
- Supervising Sound Localization by In-the-wild Egomotion (CVPR 2025 · similarity 0.88)
- Whence the Voice? Self-supervised Dual-source Audio-Visual Localisation via Selective Convergence (ECCV 2026 · similarity 0.87)
- Object-aware Sound Source Localization via Audio-Visual Scene Understanding (CVPR 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.
