The task of music structure analysis refers to automatically identifying the location and the nature of musical sections within a song. In the supervised scenario, structural annotations generally result from exhaustive data collection processes, which represents one of the main challenges of this task. Moreover, both the subjectivity of music structure and the hierarchical characteristics it exhibits make the obtained structural annotations not fully reliable, in the sense that they do not convey a “universal ground-truth” unlike other tasks in music information retrieval. On the other hand, the quickly growing quantity of available music data has enabled weakly supervised and self-supervised approaches to achieve impressive results on a wide range of music-related problems. In this work, a self-supervised learning method is proposed to learn robust multi-level music representations prior to structural segmentation using contrastive learning. To this end, sets of frames sampled at different levels of detail are used to train a deep neural network in a disentangled manner. The proposed method is evaluated on both flat and multi-level segmentation. We show that each distinct sub-region of the output embeddings can efficiently account for structural similarity at their own targeted level of detail, which ultimately improves performance of downstream flat and multi-level segmentation. Finally, complementary experiments are carried out to study how the obtained representations can be further adapted to specific datasets using a supervised fine-tuning objective in order to facilitate structure retrieval in domains where human annotations remain scarce.

An important amount of work has been devoted to the task of music classification. Despite promising results achieved by convolutional neural networks, there still exists a gap left to be filled for such models to perform well in real-world applications. In this work, we address the issue of ambiguity that can arise in many classification problems. We propose a method based on adaptive label smoothing that aims at implicitly modelling perceptual vagueness among classes to improve both training and testing performances. We assess our method using two state-of-the-art CNN architectures for audio classification on a variety of music mood and genre classification tasks. We show that the proposed strategy brings consistent improvements over the traditional approach, significantly improves generalization to external audio collections and emphasizes how crucial information carried by labels can be in an ambiguous music classification context.