Automatic tagging using deep convolutional neural networks

Abstract

We present a content-based automatic music tagging algorithm using fully convolutional neural networks (FCNs). We evaluate different architectures consisting of 2D convolutional layers and subsampling layers only. In the experiments, we measure the AUC-ROC scores of the architectures with different complexities and input types using the MagnaTagATune dataset, where a 4-layer architecture shows state-of-the-art performance with mel-spectrogram input. Furthermore, we evaluated the performances of the architectures with varying the number of layers on a larger dataset (Million Song Dataset), and found that deeper models outperformed the 4-layer architecture. The experiments show that mel-spectrogram is an effective time-frequency representation for automatic tagging and that more complex models benefit from more training data.

Accepted to ISMIR (International Society of Music Information Retrieval) Conference 2016

References in corpus (5)

• Adam: A Method for Stochastic Optimization
• Theano: new features and speed improvements
• Deep Karaoke: Extracting Vocals from Musical Mixtures Using a Convolutional Deep Neural Network
• A Deep Bag-of-Features Model for Music Auto-Tagging
• An End-to-End Neural Network for Polyphonic Piano Music Transcription

Cited by in corpus (56)

• Deep Learning for Audio Signal Processing
• Polyphonic Sound Event Detection and Localization using a Two-Stage Strategy
• Rethinking CNN Models for Audio Classification
• Sample-level Deep Convolutional Neural Networks for Music Auto-tagging Using Raw Waveforms
• Unsupervised Feature Learning Based on Deep Models for Environmental Audio Tagging
• End-to-end learning for music audio tagging at scale
• Weakly Labelled AudioSet Tagging with Attention Neural Networks
• Multi-label Music Genre Classification from Audio, Text, and Images Using Deep Features
• Multi-level Attention Model for Weakly Supervised Audio Classification
• Machine learning for music genre: multifaceted review and experimentation with audioset
• Kapre: On-GPU Audio Preprocessing Layers for a Quick Implementation of Deep Neural Network Models with Keras
• Convolutional Recurrent Neural Networks for Music Classification
• Audio Caption: Listen and Tell
• Evaluation of CNN-based Automatic Music Tagging Models
• DCASE 2018 Challenge Surrey Cross-Task convolutional neural network baseline
• Sample Mixed-Based Data Augmentation for Domestic Audio Tagging
• Music Genre Classification with Paralleling Recurrent Convolutional Neural Network
• Detecting Compressed Cleartext Traffic from Consumer Internet of Things Devices
• Representation Learning of Music Using Artist Labels
• Deep Cross-Modal Correlation Learning for Audio and Lyrics in Music Retrieval
• Music theme recognition using CNN and self-attention
• Sample-level CNN Architectures for Music Auto-tagging Using Raw Waveforms
• Large-scale weakly supervised audio classification using gated convolutional neural network
• Lyrics-Based Music Genre Classification Using a Hierarchical Attention Network
• Fully DNN-based Multi-label regression for audio tagging
• Learning to rank music tracks using triplet loss
• Efficient Convolutional Neural Network For Audio Event Detection
• LEAN: Light and Efficient Audio Classification Network
• Learning Music Sequence Representation from Text Supervision
• Timbre Analysis of Music Audio Signals with Convolutional Neural Networks
• Audio-Visual Sentiment Analysis for Learning Emotional Arcs in Movies
• Audio Set classification with attention model: A probabilistic perspective
• Artificial Musical Intelligence: A Survey
• Automatic Malware Description via Attribute Tagging and Similarity Embedding
• DLR : Toward a deep learned rhythmic representation for music content analysis
• Music Artist Classification with Convolutional Recurrent Neural Networks
• One Deep Music Representation to Rule Them All? : A comparative analysis of different representation learning strategies
• Exploiting Synchronized Lyrics And Vocal Features For Music Emotion Detection
• A Feature Learning Siamese Model for Intelligent Control of the Dynamic Range Compressor
• Revisiting the problem of audio-based hit song prediction using convolutional neural networks
• Singer Identification Using Convolutional Acoustic Motif Embeddings
• Towards Cover Song Detection with Siamese Convolutional Neural Networks
• Classification of Audio Segments in Call Center Recordings using Convolutional Recurrent Neural Networks
• Beyond Statistical Relations: Integrating Knowledge Relations into Style Correlations for Multi-Label Music Style Classification
• Instrument-Independent Dastgah Recognition of Iranian Classical Music Using AzarNet
• Audio Based Disambiguation Of Music Genre Tags
• Deep Learning for Audio Transcription on Low-Resource Datasets
• A Novel Approach for Effective Learning in Low Resourced Scenarios
• Towards Music Captioning: Generating Music Playlist Descriptions
• Multi-task Learning with Metadata for Music Mood Classification
• A joint separation-classification model for sound event detection of weakly labelled data
• Multi-scale Embedded CNN for Music Tagging (MsE-CNN)
• A Visual Domain Transfer Learning Approach for Heartbeat Sound Classification
• Unsupervised Source Separation By Steering Pretrained Music Models
• An approach to hummed-tune and song sequences matching
• A Joint Detection-Classification Model for Audio Tagging of Weakly Labelled Data