musicnn: Pre-trained Convolutional Neural Networks for Music Audio Tagging

Titlemusicnn: Pre-trained Convolutional Neural Networks for Music Audio Tagging
Publication TypeConference Paper
Year of Publication2019
Conference NameInternational Society for Music Information Retrieval (ISMIR)
AuthorsPons, J., & Serra X.
Conference Start Date04/11/2019
Conference LocationDelft, Netherlands
AbstractPronounced as "musician", the musicnn library contains a set of pre-trained musically motivated convolutional neural networks for music audio tagging: this https URL. This repository also includes some pre-trained vgg-like baselines. These models can be used as out-of-the-box music audio taggers, as music feature extractors, or as pre-trained models for transfer learning. We also provide the code to train the aforementioned models: this https URL. This framework also allows implementing novel models. For example, a musically motivated convolutional neural network with an attention-based output layer (instead of the temporal pooling layer) can achieve state-of-the-art results for music audio tagging: 90.77 ROC-AUC / 38.61 PR-AUC on the MagnaTagATune dataset --- and 88.81 ROC-AUC / 31.51 PR-AUC on the Million Song Dataset.
preprint/postprint documenthttps://arxiv.org/abs/1909.06654
Additional material: 
intranet