PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition
Abstract
Audio pattern recognition is an important research topic in the machine learning area, and includes several tasks such as audio tagging, acoustic scene classification, music classification, speech emotion classification and sound event detection. Recently, neural networks have been applied to tackle audio pattern recognition problems. However, previous systems are built on specific datasets with limited durations. Recently, in computer vision and natural language processing, systems pretrained on large-scale datasets have generalized well to several tasks. However, there is limited research on pretraining systems on large-scale datasets for audio pattern recognition. In this paper, we propose pretrained audio neural networks (PANNs) trained on the large-scale AudioSet dataset. These PANNs are transferred to other audio related tasks. We investigate the performance and computational complexity of PANNs modeled by a variety of convolutional neural networks. We propose an architecture called Wavegram-Logmel-CNN using both log-mel spectrogram and waveform as input feature. Our best PANN system achieves a state-of-the-art mean average precision (mAP) of 0.439 on AudioSet tagging, outperforming the best previous system of 0.392. We transfer PANNs to six audio pattern recognition tasks, and demonstrate state-of-the-art performance in several of those tasks. We have released the source code and pretrained models of PANNs: https://github.com/qiuqiangkong/audioset_tagging_cnn.
The following analysis is generated by AI and may contain errors. It is separate from the authors’ abstract and is not peer review.
Research summary
This paper introduces Pretrained Audio Neural Networks (PANNs) that leverage large-scale datasets to achieve state-of-the-art performance in audio pattern recognition tasks. The innovative methodologies and comprehensive evaluations presented in this work significantly contribute to the advancement of audio machine learning, particularly in the context of transfer learning and model efficiency.
Methodology
The paper presents a novel approach to audio pattern recognition through the introduction of Pretrained Audio Neural Networks (PANNs) trained on the large-scale AudioSet dataset. The methodology includes the innovative Wavegram-Logmel-CNN architecture that integrates both log-mel spectrograms and waveform inputs, which is a significant advancement over traditional methods that typically rely solely on one type of input feature. The paper also discusses various convolutional neural network architectures, including adaptations of ResNets and MobileNets, and emphasizes the importance of data balancing and augmentation techniques in improving model performance.
Results & evaluation
The experimental evaluation is robust, demonstrating the effectiveness of PANNs across multiple audio pattern recognition tasks. The authors report state-of-the-art performance metrics, including a mean average precision (mAP) of 0.439 on AudioSet tagging, surpassing previous benchmarks. The experiments are well-structured, comparing various architectures and configurations, and they include a comprehensive analysis of class-wise performance, which highlights the strengths and weaknesses of the proposed systems.
Reproducibility
The authors have made their source code and pretrained models publicly available, which enhances the reproducibility of their results. However, the paper could benefit from more detailed descriptions of the training procedures and hyperparameter settings to ensure that other researchers can replicate the experiments accurately.
Limitations
One limitation of the study is the reliance on a single large dataset (AudioSet) for training, which may affect the generalizability of the models to other audio domains. Additionally, while the paper addresses computational complexity, there is limited discussion on the practical deployment of these models in real-world applications, particularly on resource-constrained devices.
Broader impact
The work has significant implications for various applications in audio analysis, including music classification, environmental sound recognition, and speech emotion classification. The advancements in transfer learning for audio tasks could facilitate the development of more efficient models that require less labeled data, which is a critical challenge in the field. This paper introduces Pretrained Audio Neural Networks (PANNs) that leverage large-scale datasets to achieve state-of-the-art performance in audio pattern recognition tasks. The innovative methodologies and comprehensive evaluations presented in this work significantly contribute to the advancement of audio machine learning, particularly in the context of transfer learning and model efficiency.
Curator note
Kong et al., QMUL; pretrained CNNs on AudioSet; became the standard backbone for audio tagging and classification
Author of this paper? Send a correction.