Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866911441827659776 |
|---|---|
| author | Yuan, Kuang Gao, Yang Li, Xilin Mei, Xinhao Zadissa, Syavosh Pruthi, Tarun Sereshki, Saeed Bagheri |
| author_facet | Yuan, Kuang Gao, Yang Li, Xilin Mei, Xinhao Zadissa, Syavosh Pruthi, Tarun Sereshki, Saeed Bagheri |
| contents | Acoustic scene classification (ASC) models on edge devices typically operate under fixed class assumptions, lacking the transferability needed for real-world applications that require adaptation to new or refined acoustic categories. We propose ContrastASC, which learns generalizable acoustic scene representations by structuring the embedding space to preserve semantic relationships between scenes, enabling adaptation to unseen categories without retraining. Our approach combines supervised contrastive fine-tuning of pre-trained models with contrastive representation distillation to transfer this structured knowledge to compact student models. Our evaluation shows that ContrastASC demonstrates improved few-shot adaptation to unseen categories while maintaining strong closed-set performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_03728 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation Yuan, Kuang Gao, Yang Li, Xilin Mei, Xinhao Zadissa, Syavosh Pruthi, Tarun Sereshki, Saeed Bagheri Sound Machine Learning Audio and Speech Processing Signal Processing Acoustic scene classification (ASC) models on edge devices typically operate under fixed class assumptions, lacking the transferability needed for real-world applications that require adaptation to new or refined acoustic categories. We propose ContrastASC, which learns generalizable acoustic scene representations by structuring the embedding space to preserve semantic relationships between scenes, enabling adaptation to unseen categories without retraining. Our approach combines supervised contrastive fine-tuning of pre-trained models with contrastive representation distillation to transfer this structured knowledge to compact student models. Our evaluation shows that ContrastASC demonstrates improved few-shot adaptation to unseen categories while maintaining strong closed-set performance. |
| title | Lightweight and Generalizable Acoustic Scene Representations via Contrastive Fine-Tuning and Distillation |
| topic | Sound Machine Learning Audio and Speech Processing Signal Processing |
| url | https://arxiv.org/abs/2510.03728 |