Research map / Vision-language and multimodal / Cross-modal learning
Enhancing Sound Source Localization via False Negative Elimination
Zengjie Song, Jiangshe Zhang, Yuxi Wang, Junsong Fan, Zhaoxiang Zhang
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 stop-gradient (ablation) | SSPL from scratch · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.141 -> 0.570 | Improved |
| add ImageNet pre-training (ablation) | SSPL with stop-gradient · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.570 -> 0.671 | Improved |
| add predictive coding module with 5 recursive cycles (ablation) | SSPL w/o PCM · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.671 -> 0.743 | Improved |
| use 3 PCM cycles instead of 1 (ablation) | SSPL w/ PCM · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.655 -> 0.719 | Improved |
| use min-max normalization instead of softmax scaling (ablation) | SSPL attention module · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.667 -> 0.671 | Improved |
| use min-max normalization instead of ReLU scaling (ablation) | SSPL attention module · SoundNet-Flickr (SSPL ablation, 10k) | cIoU: 0.353 -> 0.671 | Improved |
| add fine-tuning of both backbones (ablation) | SACL baseline · SoundNet-Flickr (SACL ablation, 10k) | cIoU: 0.619 -> 0.707 | Improved |
| add pseudo masks (ablation) | SACL fine-tuned · SoundNet-Flickr (SACL ablation, 10k) | cIoU: 0.707 -> 0.755 | Improved |
| add false negative detection (ablation) | SACL fine-tuned · SoundNet-Flickr (SACL ablation, 10k) | cIoU: 0.707 -> 0.763 | Improved |
| add false negative detection on top of pseudo masks (ablation) | SACL with pseudo mask · SoundNet-Flickr (SACL ablation, 10k) | cIoU: 0.755 -> 0.815 | Improved |
| add PCM to SSPL | SSPL w/o PCM · VGG-SS (trained on VGG-Sound144k) | cIoU: 0.270 -> 0.339 | Improved |
| replace TIE with SACL | TIE · VGG-SS (trained on VGG-Sound144k) | cIoU: 0.386 -> 0.380 | Hurt |
Closest papers by meaning
- Learning Audio-Visual Source Localization via False Negative Aware Contrastive Learning (CVPR 2023 · similarity 0.94)
- Robust Audio-Visual Contrastive Learning for Proposal-Based Self-Supervised Sound Source Localization in Videos (TPAMI 2024 · similarity 0.94)
- Improving Sound Source Localization with Joint Slot Attention on Image and Audio (CVPR 2025 · similarity 0.91)
- Object-aware Sound Source Localization via Audio-Visual Scene Understanding (CVPR 2025 · similarity 0.91)
- T-VSL: Text-Guided Visual Sound Source Localization in Mixtures (CVPR 2024 · similarity 0.89)
- Prompting Segmentation with Sound Is Generalizable Audio-Visual Source Localizer (AAAI 2024 · similarity 0.89)
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.
