Research map / Label-efficient and robust learning / Self-supervised and contrastive
A Sequential Self Teaching Approach for Improving Generalization in Sound Event Recognition
Anurag Kumar, Vamsi Ithapu
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 |
|---|---|---|---|
| Use sequential co-supervision from previous models | WEANET · AudioSet | mAP: 0.366 -> 0.398 | Improved |
| Use two previous-stage teachers instead of one (ablation) | SUSTAIN on WEANET · AudioSet | mAP: 0.394 -> 0.398 | Improved |
| Combine ground-truth labels with teacher predictions rather than using either alone (ablation) | SUSTAIN on WEANET · AudioSet | mAP: up to 5.7% | Improved |
| Increase the number of sequential stages on the noisy training set (ablation) | SUSTAIN with WEANET^L · FSDKaggle Noisy | lwlrap: 0.436 -> 0.472 | Improved |
| Continue co-supervision beyond one stage on clean curated data (ablation) | SUSTAIN with WEANET^L · FSDKaggle Curated | lwlrap | Hurt |
| Use class-specific attention instead of average pooling (ablation) | WEANET · AudioSet | mAP: 0.352 -> 0.366 | Improved |
| Transfer features from a SUSTAIN-trained model rather than the base model | AudioSet-trained WEANET · ESC-50 transfer | accuracy: 92.6 -> 94.1 | Improved |
| Transfer features from a SUSTAIN-trained model rather than the base model | AudioSet-trained WEANET · FSDKaggle Noisy transfer | lwlrap: 0.486 -> 0.503 | Improved |
| Transfer features from a SUSTAIN-trained model rather than the base model | AudioSet-trained WEANET · FSDKaggle Curated transfer | lwlrap: 0.712 -> 0.728 | Improved |
Closest papers by meaning
- MERT: Acoustic Music Understanding Model with Large-Scale Self-supervised Training (ICLR 2024 · similarity 0.87)
- Enhanced Audio Tagging via Multi- to Single-Modal Teacher-Student Mutual Learning (AAAI 2021 · similarity 0.85)
- Unified Speech Recognition: A Single Model for Auditory, Visual, and Audiovisual Inputs (NeurIPS 2024 · similarity 0.83)
- SSLAM: Enhancing Self-Supervised Models with Audio Mixtures for Polyphonic Soundscapes (ICLR 2025 · similarity 0.83)
- DiffSED: Sound Event Detection with Denoising Diffusion (AAAI 2024 · similarity 0.83)
- SALMONN: Towards Generic Hearing Abilities for Large Language Models (ICLR 2024 · similarity 0.82)
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.
