Research map / Video / Video segmentation and tracking
Advancing Weakly-Supervised Audio-Visual Video Parsing via Segment-Wise Pseudo Labeling
Jinxing Zhou, Dan Guo, Yiran Zhong, Meng Wang
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 PLG video-level visual pseudo labels to HAN (ablation) | HAN · LLP test, visual segment-level | V F-score: 52.9 -> 64.1 | Improved |
| add PLG video-level pseudo labels to HAN, audio effect (ablation) | HAN · LLP test, audio segment-level | A F-score: 60.1 -> 59.8 | No clear effect |
| add naive segment-level BCE loss (PLE) on PLG labels (ablation) | HAN + PLG · LLP test, segment-level | Type@AV F-score: 60.5 -> 61.6 | Improved |
| use richness-aware loss instead of naive BCE segment loss (ablation) | HAN + PLG · LLP test, segment-level | Type@AV F-score: 61.6 -> 62.0 | Improved |
| add category-richness alignment to the richness-aware loss (ablation) | HAN + PLG · LLP test, segment-level | Event@AV F-score: 58.3 -> 60.2 | Improved |
| add segment-richness alignment to the richness-aware loss (ablation) | HAN + PLG · LLP test, segment-level | Type@AV F-score: 60.5 -> 61.3 | Improved |
| add PLD denoising of visual pseudo labels (ablation) | HAN + PLG + PLE · LLP test, segment-level | Type@AV F-score: 62.0 -> 63.1 | Improved |
| add PLD to visual pseudo labels (K=5, alpha=30) (ablation) | PLG visual pseudo labels · LLP val, pseudo-label quality | average F-score: 67.95 -> 70.63 | Improved |
| add PLD to audio pseudo labels (K=6, alpha=400) (ablation) | PLG audio pseudo labels · LLP val, pseudo-label quality | average F-score: 75.93 -> 76.06 | No clear effect |
| raise visual threshold tau_v from 0.041 to 0.042 (ablation) | PLG visual pseudo labels · LLP val | segment-level V F-score: 71.08 -> 51.51 | Hurt |
| use prompt VP1 rather than VP3 for CLIP (ablation) | PLG visual pseudo labels · LLP val | segment-level V F-score: 67.60 -> 71.08 | Improved |
| use PLG pseudo labels instead of assigning video labels to the visual modality (ablation) | visual pseudo labels · LLP val | video-level precision: 66.96 -> 86.88 | Improved |
Closest papers by meaning
- Weakly-Supervised Audio-Visual Video Parsing with Prototype-based Pseudo-Labeling (CVPR 2024 · similarity 0.96)
- UWAV: Uncertainty-weighted Weakly-supervised Audio-Visual Video Parsing (CVPR 2025 · similarity 0.92)
- Multimodal Class-aware Semantic Enhancement Network for Audio-Visual Video Parsing (AAAI 2025 · similarity 0.89)
- Dense Audio-Visual Event Localization Under Cross-Modal Consistency and Multi-Temporal Granularity Collaboration (AAAI 2025 · similarity 0.88)
- Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction (AAAI 2025 · similarity 0.88)
- PreFM: Online Audio-Visual Event Parsing via Predictive Future Modeling (NeurIPS 2025 · similarity 0.88)
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.
