Research map / Detection and segmentation / Open-set and few-shot detection
Discriminative Sounding Objects Localization via Self-supervised Audiovisual Matching
Di Hu, Rui Qian, Minyue Jiang, Xiao Tan, Shilei Wen, Errui Ding, Weiyao Lin, Dejing Dou
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 alternating localization/classification training with K-means pseudo-labels in stage 1 (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU@0.3: 17.4 -> 32.3 | Improved |
| add product operation (sounding-area filter) in stage 2 (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU: 2.6 -> 32.3 | Improved |
| add audiovisual consistency loss Lc in stage 2 (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU: 18.0 -> 32.3 | Improved |
| add audiovisual consistency loss Lc in stage 2 (ablation) | two-stage sounding object localization · MUSIC-Duet | CIoU: 22.4 -> 30.2 | Improved |
| add localization loss L1 alongside Lc (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU: 0.0 -> 2.6 | Improved |
| use ImageNet-pretrained visual backbone instead of scratch (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU@0.3: 7.8 -> 32.3 | Improved |
| use ImageNet-pretrained visual backbone instead of scratch (ablation) | two-stage sounding object localization · MUSIC-solo | IoU@0.5: 25.0 -> 51.4 | Improved |
| lambda in [0.5, 1.0] versus 0.3 or 1.5 (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU@0.3: 11.0 -> 32.3 | Improved |
| increase number of clusters from 11 to 20 (or 13) (ablation) | two-stage sounding object localization · MUSIC-Synthetic | CIoU@0.3: 32.3 -> 29.5 | No clear effect |
| change mask threshold from 0.05 to 0.07 (ablation) | two-stage sounding object localization · MUSIC-solo | IoU@0.5: 51.4 -> 43.1 | Hurt |
| class-aware distribution matching vs Sound-of-pixel | two-stage sounding object localization · MUSIC-Synthetic | CIoU@0.3: 8.1 -> 32.3 | Improved |
| class-aware method vs Sound-of-pixel | two-stage sounding object localization · AudioSet-instrument-multi | NSA: 88.8 -> 56.8 | Hurt |
Closest papers by meaning
- Class-Aware Sounding Objects Localization via Audiovisual Correspondence (TPAMI 2021 · similarity 0.94)
- A Proposal-Based Paradigm for Self-Supervised Sound Source Localization in Videos (CVPR 2022 · similarity 0.90)
- Self-Supervised Object Detection From Audio-Visual Correspondence (CVPR 2022 · similarity 0.89)
- Object-aware Sound Source Localization via Audio-Visual Scene Understanding (CVPR 2025 · similarity 0.87)
- Learning to Visually Localize Sound Sources from Mixtures without Prior Source Knowledge (CVPR 2024 · similarity 0.87)
- Prompting Segmentation with Sound Is Generalizable Audio-Visual Source Localizer (AAAI 2024 · 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.
