Research map / Label-efficient and robust learning / Semi-supervised learning
Learning to Localize Sound Source in Visual Scenes
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Increase unsupervised training data from 10k to 144k pairs (ablation) | Unsupervised sound localization model · Sound Source Localization dataset | cIoU: 43.6 -> 66.0 | Helped |
| Add 2.5k supervised samples to 10k unsupervised samples (ablation) | Sound localization model · Sound Source Localization dataset | cIoU: 43.6 -> 82.8 | Helped |
| Combine 2.5k supervised samples with 10k unsupervised samples instead of using 2.5k supervised samples alone (ablation) | Sound localization model · Sound Source Localization dataset | cIoU: 80.4 -> 82.8 | Helped |
| Use ReLU before softmax attention (ablation) | Unsupervised model trained on 144k samples · Sound Source Localization dataset | cIoU: 66.0 -> 52.4 | Hurt |
| Use ReLU before softmax attention (ablation) | Sup. 2.5k + Unsup. 10k model · Sound Source Localization dataset | cIoU: 82.8 -> 84.0 | Helped |
| Use ReLU before softmax attention (ablation) | Sup. 2.5k + Unsup. 10k model · Sound Source Localization dataset | AUC: 62.0 -> 61.9 | No clear effect |
| Add 0.5k supervised samples to 10k unsupervised samples (ablation) | Sound localization model · Sound Source Localization dataset | cIoU: 43.6 -> 78.0 | Helped |
| Increase supervised samples from 0.5k to 1.0k while retaining 10k unsupervised samples (ablation) | Semi-supervised sound localization model · Sound Source Localization dataset | cIoU: 78.0 -> 82.4 | Helped |
| Increase supervised samples from 1.0k to 1.5k while retaining 10k unsupervised samples (ablation) | Semi-supervised sound localization model · Sound Source Localization dataset | cIoU: 82.4 -> 82.0 | No clear effect |
| Increase supervised samples from 1.5k to 2.5k while retaining 10k unsupervised samples (ablation) | Semi-supervised sound localization model · Sound Source Localization dataset | AUC: 61.3 -> 62.0 | Helped |
| Use ReLU attention with 0.5k supervised samples and 10k unsupervised samples (ablation) | Semi-supervised sound localization model · Sound Source Localization dataset | cIoU: 78.0 -> 79.2 | Helped |
| Use ReLU attention with 1.5k supervised samples and 10k unsupervised samples (ablation) | Semi-supervised sound localization model · Sound Source Localization dataset | AUC: 61.3 -> 61.8 | Helped |
Closest papers by meaning
- Object-aware Sound Source Localization via Audio-Visual Scene Understanding (CVPR 2025 · similarity 0.89)
- Improving Sound Source Localization with Joint Slot Attention on Image and Audio (CVPR 2025 · similarity 0.89)
- Prompting Segmentation with Sound Is Generalizable Audio-Visual Source Localizer (AAAI 2024 · similarity 0.88)
- T-VSL: Text-Guided Visual Sound Source Localization in Mixtures (CVPR 2024 · similarity 0.88)
- Learning to Visually Localize Sound Sources from Mixtures without Prior Source Knowledge (CVPR 2024 · similarity 0.88)
- Seeing Speech and Sound: Distinguishing and Locating Audio Sources in Visual Scenes (CVPR 2025 · similarity 0.87)
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.