通过联合软标签与特征蒸馏，实现低复杂度音频场景分类的高效模型压缩。
Joint Feature and Output Distillation for Low-complexity Acoustic Scene Classification
- 双层蒸馏：同时传递教师模型的软输出和中间特征。
- 在TAU数据集上达到59.30%准确率，优于轻量级学生模型基准。
- 适合资源受限设备部署，适用于移动端音频场景识别任务。
本文针对DCASE2025 Task 1中的低复杂度音频场景分类（ASC），提出一种多教师引导的双层知识蒸馏框架。通过预训练PaSST和CP-ResNet作为教师模型，将教师输出的软标签平均生成软目标，同时选取一个CP-ResNet进行特征级蒸馏。该方法使紧凑的学生模型（CP-Mobile）同时学习到语义分布与结构信息。在TAU Urban Acoustic Scenes 2022 Mobile数据集（开发集）上的实验表明，所提交系统最高达到59.30%的准确率。
原文摘要 · Abstract (English)
This report presents a dual-level knowledge distillation framework with multi-teacher guidance for low-complexity acoustic scene classification (ASC) in DCASE2025 Task 1. We propose a distillation strategy that jointly transfers both soft logits and intermediate feature representations. Specifically, we pre-trained PaSST and CP-ResNet models as teacher models. Logits from teachers are averaged to generate soft targets, while one CP-ResNet is selected for feature-level distillation. This enables the compact student model (CP-Mobile) to capture both semantic distribution and structural information from teacher guidance. Experiments on the TAU Urban Acoustic Scenes 2022 Mobile dataset (development set) demonstrate that our submitted systems achieve up to 59.30\% accuracy.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。