# A Two-Stage Approach to Device-Robust Acoustic Scene Classification

arXiv: https://arxiv.org/abs/2011.01447 · HTML: https://mlsanity.com/paper/2011.01447/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 3 Nov 2020.

## Built on by (most cited first, 9 of 9)

- [Voice2Series: Reprogramming Acoustic Models for Time Series Classification](https://mlsanity.com/paper/2106.09296/) (2021)
- [A Lottery Ticket Hypothesis Framework for Low-Complexity Device-Robust Neural Acoustic Scene Classification](https://mlsanity.com/paper/2107.01461/) (2021)
- [A Variational Bayesian Approach to Learning Latent Variables for Acoustic Knowledge Transfer](https://mlsanity.com/paper/2110.08598/) (2021)
- [Attentive max feature map and joint training for acoustic scene classification](https://mlsanity.com/paper/2104.07213/) (2021)
- [Adversarial Domain Adaptation with Paired Examples for Acoustic Scene Classification on Different Recording Devices](https://mlsanity.com/paper/2110.09598/) (2021)
- [Robust, General, and Low Complexity Acoustic Scene Classification Systems and An Effective Visualization for Presenting a Sound Scene Context](https://mlsanity.com/paper/2210.08610/) (2022)
- [A study on joint modeling and data augmentation of multi-modalities for audio-visual scene classification](https://mlsanity.com/paper/2203.04114/) (2022)
- [Improving Acoustic Scene Classification with City Features](https://mlsanity.com/paper/2503.16862/) (2025)
- [ASCMamba: Multimodal Time-Frequency Mamba for Acoustic Scene Classification](https://mlsanity.com/paper/2508.15632/) (2025)

## Abstract (arXiv)

To improve device robustness, a highly desirable key feature of a competitive data-driven acoustic scene classification (ASC) system, a novel two-stage system based on fully convolutional neural networks (CNNs) is proposed. Our two-stage system leverages on an ad-hoc score combination based on two CNN classifiers: (i) the first CNN classifies acoustic inputs into one of three broad classes, and (ii) the second CNN classifies the same inputs into one of ten finer-grained classes. Three different CNN architectures are explored to implement the two-stage classifiers, and a frequency sub-sampling scheme is investigated. Moreover, novel data augmentation schemes for ASC are also investigated. Evaluated on DCASE 2020 Task 1a, our results show that the proposed ASC system attains a state-of-the-art accuracy on the development set, where our best system, a two-stage fusion of CNN ensembles, delivers a 81.9% average accuracy among multi-device test data, and it obtains a significant improvement on unseen devices. Finally, neural saliency analysis with class activation mapping (CAM) gives new insights on the patterns learnt by our models.