Transfer learning for music classification and regression tasks

Abstract

In this paper, we present a transfer learning approach for music classification and regression tasks. We propose to use a pre-trained convnet feature, a concatenated feature vector using the activations of feature maps of multiple layers in a trained convolutional network. We show how this convnet feature can serve as general-purpose music representation. In the experiments, a convnet is trained for music tagging and then transferred to other music-related classification and regression tasks. The convnet feature outperforms the baseline MFCC feature in all the considered tasks and several previous approaches that are aggregating MFCCs as well as low- and high-level music features.

18th International Society of Music Information Retrieval (ISMIR) Conference, Suzhou, China, 2017

References in corpus (7)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• SoundNet: Learning Sound Representations from Unlabeled Video
• Kapre: On-GPU Audio Preprocessing Layers for a Quick Implementation of Deep Neural Network Models with Keras
• Look, Listen and Learn
• Explaining Deep Convolutional Neural Networks on Music Classification
• Towards Playlist Generation Algorithms Using RNNs Trained on Within-Track Transitions
• The Effects of Noisy Labels on Deep Convolutional Neural Networks for Music Tagging

Cited by in corpus (17)

• Rethinking CNN Models for Audio Classification
• Machine learning for music genre: multifaceted review and experimentation with audioset
• Fine-tuning deep learning model parameters for improved super-resolution of dynamic MRI with prior-knowledge
• An Interdisciplinary Review of Music Performance Analysis
• Texture Selection for Automatic Music Genre Classification
• Comparison and Analysis of Deep Audio Embeddings for Music Emotion Recognition
• Bottom-up Broadcast Neural Network For Music Genre Classification
• MusiCoder: A Universal Music-Acoustic Encoder Based on Transformers
• Metric Learning vs Classification for Disentangled Music Representation Learning
• Transfer Learning in Spatial-Temporal Forecasting of the Solar Magnetic Field
• JEPOO: Highly Accurate Joint Estimation of Pitch, Onset and Offset for Music Information Retrieval
• Towards an efficient deep learning model for musical onset detection
• Complex Network-Based Approach for Feature Extraction and Classification of Musical Genres
• Detection and classification of vocal productions in large scale audio recordings
• Towards noise robust trigger-word detection with contrastive learning pre-task for fast on-boarding of new trigger-words
• MS-SincResNet: Joint learning of 1D and 2D kernels using multi-scale SincNet and ResNet for music genre classification
• Automatic Realistic Music Video Generation from Segments of Youtube Videos